Site Preparation and Installation Document Set 3 Hard Copy Documents
Go to http://www.intervoice.com/support/techinfo/ Select Site Preparation Information. (no password required) View the list to determine which document applies to your system.
Because Intervoice primarily provides its product information as online documents on product CD-ROMs and the Web, only a limited number of documents are available as hard copy for shipments. These include product quick reference cards for hardware, system installation, as well as vendor partner products. Contact your Sales Representative for more information.
Related Intervoice Documents
The following documents contain useful information about your product. The documents are listed in functional order.

InterSoft Version 3

Loading InterSoft Software (InterSoft version 3) Doc Number 60001263
Describes how to install the latest InterSoft version 3 software. It also describes how to edit configuration files and apply system utilities.
InterSoft System Administration (InterSoft version 3) Doc Number 60001092
Explains how to prepare a system for the runtime production environment. It also provides commands and explanations of the basic system administration functions of an InterSoft version 3 system, including starting and stopping the runtime system, monitoring system status, creating and editing the various application tables, and working with voice messages.
InterSoft System Utilities (InterSoft version 3) Doc Number 60001094
Describes the system utilities used to prepare, operate, and monitor a system in runtime.
Troubleshooting InterSoft Systems (InterSoft version 3) Doc Number 60001078
Explains how to isolate problems on an InterSoft version 3 runtime system and troubleshoot once a problem has been isolated. It describes the various tools available for problem isolation, explains how to use each tool (or what document to reference for instructions), and explains how to analyze the results of the diagnostic tests.

Sound Forge

Sound Forge Manual
Provides overview and installation information. It describes how to use and optimize Sound Forge. See the product CD-ROM and Web site: http://mediasoftware.sonypictures.com/download/.

Conventions

This document uses the following keyboard, command prompt, and menu conventions:

Example

<Enter> <Ctrl><Esc> [d:\appl]dir <Enter>

Instructs you to.

Press the Enter key. Hold the Control key while pressing the Escape key. Enter a command at a Windows command prompt.
Note These commands are not case-sensitive.
Edit > Paste > Save
Select Edit from the menu bar, choose Paste, and then select Save from its menu.

Insert the Y-cable connectors into the sound card connectors in your computer.
Input Jack to Preamplifier Audio Cable Y-Cable to Speakers Output Jack
Use the manufacturers instructions to install the other components.
Installing the Sonic Foundry Batch Converter
If Batch Converter 5.0 is not on the Sound Forge CD, Sound Forge 6.0 and 7.0 customers with a valid serial number can download Batch Converter 5.0 from this Web site URL:
http://mediasoftware.sonypictures.com/download/step2.asp?DID=339& #Return/. It is important that you register your copy of Sound Forge and
the Batch Converter to obtain technical support.
Note Sound Forge 7.0 users should contact Intervoice
Customer Support if the serial number supplied with their copy of Sound Forge 7.0 fails to enable the Batch Converter 5.0. Periodic updates are available to Sound Forge and the Batch Converter. Intervoice recommends checking the Web site first for the latest updates to the Batch Converter. For more Sound Forge information, contact technical support at http://mediasoftware.sonypictures.com/support. For Intervoice 24-hour support, contact the following support office: (outside USA) Tel: +955-4688 or +454-8130 (within USA) Tel: (800) 955-4688 or (972) 454-8130 To install the Batch Converter:
Run the Batch Converter installation file. For example, if it was downloaded to your desktop, double-click it. Or, if it is on a CDROM, go to the directory and start the.exe file. Follow the installation screens. After installing the Batch Converter, download the Intervoice Example Batch Script file at: www.intervoice.com/support/techinfo/voicerecording. Save the IVBExampleBatchScriptV6.zip file to the Batch Converter directory. Unzip the batch script, IVBExampleBatchScriptV6.bcs. Refer to the topic Using the Sample Batch Script on page 5-3 for more information.

Installing CRVoice2

Numerical file extensions are not recognized correctly by Windows, so Intervoice created a proprietary extension,.ivc, to help identify Intervoice files on Windows systems. However, when.ivc files are transferred to an Intervoice InterSoft IQTalk system, the.ivc extensions (for example BankABC502.ivc) must be replaced with numeric extensions such as m.502. The CRVoice2 utility renames the.ivc files to m.### files. To install CRVoice2:
Download the file from the Intervoice Web site at:
www.intervoice.com/support/techinfo/voicerecording.
Save the file to the directory where you will be converting your.wav master files to.ivc files for InterSoft IQTalk systems.
Refer to Using the CRVoice2 Utility on page 6-4 for more information.

Installing Sound Forge

To install Sound Forge:
Ensure that the Batch Converter, sample Intervoice script, and CRVoice2 utility have already been installed. Insert the Sound Forge CD-ROM in the CD-ROM drive. The installation program should autostart. Follow the installation instruction screens. Go to the Web site and download the latest update to Sound Forge:

Avoiding Clipping

Sound Forge clips a recording when it exceeds the maximum storable amplitude or when any value beyond the maximum is cut off, resulting in audible distortion. Clipping destroys a recordings quality; therefore, clipped recordings must be rerecorded. Clipping occurs when the level meters are pushed into the red zone (see Figure 2-3 on page 2-11). Figure 2-2 illustrates a voice recording without clipping.

No clipping

Figure 2-2. Voice Recording Without Clipping
To prevent clipping, set the recording input level just low enough to enable the voice talent to speak at normal volume without pushing the level meters into the red zone.
The flat tops in Figure 2-3 illustrate clipping.

Clipping

Figure 2-3. Voice Recording with Clipping
Performing the Noise Floor Test
Always check the background noise and a noise floor of the recording studio and equipment before starting a recording session. Recordings made with high background noise will not provide a clear voice sound during playback. Sound Forge must be installed to perform this test. A high noise floor is a common cause of poor audio quality and clicks in compressed files. (See Using Sound Forge to Record Messages, Chapter 4 for more information.) To check for noise using Sound Forge tools, follow these steps:
Ensure that the microphone levels on the preamplifier are set correctly for a recording session, that is, as loud as possible without causing clipping (see Setting Microphone Levels on page 2-8). Verify that the recording attributes are 11,025 Hz (or 11.025 KHz), 16 bit, Mono. Record five seconds of silence, at the standard volume level. Do not speak while recording. The recording studio should be as quiet as possible during this test. Stop recording. Close the Record dialog box. With a sound file window open, go to the main Sound Forge window and select Process > DC Offset. The DC Offset <filename> window appears.
In Sound Forge 6, set the Preset field to Intervoice DC Offset Removal. In Sound Forge 7, set the Preset field to Remove DC offset.
Select the Automatically detect and remove option to remove the DC offset, and click OK. To view statistics, select Tools > Statistics. The Statistics window appears.
Verify that the RMS power in the statistics dialog box is below 60 dB (for example, 65 dB). Verify that the Average value (DC offset) statistic is 0.

Table 3-1 charts the differences between the InterSoft master format and the standard runtime format.
Table 3-1. Format Differences Format Name
Windows Master Format InterSoft IQTalk Runtime Format VoiceXML Runtime Format

Sample Rate

11,025 smp/sec 8,000 smp/sec

Sample Size (Mono)

16 bits/sample

Data Format

Linear.wav

Size of 10 sec Message

220 KB

4 bits/sample

ADPCM (.ivc) G.721

8,000 smp/sec

8 bits/sample

A-law or -law PCM

After messages are recorded, they must be compressed from master.wav files into ADPCM (.ivc) or A-law or -law PCM.wav files. Sound Forge provides a Batch Converter that enables you to easily modify, edit, and convert large groups of messages. See Batch Converter Overview on page 5-2 for more information.
Setting Up Automatic Labeling, 4-2 Setting Recording Attributes, 4-5 Recording Process Checklist, 4-6 Recording Methods, 4-7 Recording Messages, 4-9 Reviewing Messages, 4-17 Removing Intervoice Batch Scripts, 4-18
Using Sound Forge to Record Messages
This chapter explains how to set up the automatic labeling functions and describes the voice recording process, the methods of recording voice messages, and the use of Sound Forge to record voice messages.

Sample Screens

This document contains representations of application screens. These screens are used as examples only and may not appear exactly as those on your system.
Setting Up Automatic Labeling
Windows 2000 allows long filenames. Assigning long filenames (BANKABC.1stGreeting.1048, for example) to message files may help keep track of messages and associate them with the correct applications. Long filenames make it easier to identify message files to be renamed for use in runtime. (See Filename Formats (.ivc to m.###) on page 6-2 and.) The automatic labeling function provided by Sound Forge can be used to assign descriptive filenames to master voice files. It is also used to automatically name and number regions or files during the recording process. Intervoice recommends that you set up the automatic labeling function before recording messages.

Converting the.wav Files to Runtime Format Overview
A series of transformational steps is required to convert a.wav file into the runtime file format. All of these steps can be automated into a single batch process in the Batch Converter. The Batch Converter can apply this process to a large number of master.wav files in a single step, creating a set of runtime files, one for each master.wav file. To convert a master.wav file into a runtime file, the Batch Converter must perform a specific set of processes in a specific order on the original.wav master file. In version 5.0, Sonic Foundry calls these steps plug-ins (presets in Sound Forge 4.5). The final script is called a plugin chain. The basic recording process provides a clean audio signal digitized to the systems hard drive in.wav file format. However, there are several other processing steps that must be applied to an audio file after it is recorded, but before it can be placed on an Intervoice system. The Intervoice sample Batch Converter postprocessing procedures are described below:

DC Offset removal

Removes any DC offset before messages are compressed. Most sound cards have some DC offset that can affect the compression process.
Front and Back silence trimming with 20-ms fades
Removes excess silence on the front and back of messages to tighten the message delivery. Generally the messages are trimmed to 20 milliseconds or less of silence on both ends. Noise at the beginning or end of messages (such as mouse or keyboard clicks) causes the auto-trimming process to trim too early or too late, leaving extra silence in the message. These messages must be hand-trimmed.
Front and Back silence buffering with 50-ms buffers
Prevents the destabilization of the ADPCM decoder when two messages are played back to back. Once the arbitrary silences are trimmed from the message, this step places a standard 50-millisecond silence buffer on the beginning and end of every message.
Message level normalization
Converts all the messages in a single application to a single standard volume level, so that listeners hear every message at the same volume.

8 KHz resampling

Converts the 11.025 KHz sample rate of the master file to an 8 KHz sampling rate.

Equalization

Performs the frequency manipulation required to make high-quality message playback over telephone lines.
ADPCM compression for A-law playback and Intervoice formatting
Compresses the linear format to ADPCM to reduce the size of the message files on the Intervoice system. This step is where you specify whether a message is to be played as -law or A-law, or to be compressed to 32 KB.
Using the Batch Converter

To use the Batch Converter, follow these steps:
From the main Sonic Foundry Batch Converter window, select File > Open Script. Navigate to and open the Intervoice Example Batch Script, IVBExampleBatchScriptV6.bcs. The Sonic Foundry Batch Converter IVBExampleBatchScriptV6.bcs window appears in the Script tab view.
Do not select any of the options on the window. Select the directory where you want the output files saved. In the File format area, set the Format and Template.
Note If you change any selection, save the changes as
a new batch script and retain the original sample batch script.
To choose the files that are to be converted, go to the main Sonic Foundry Batch Converter window and select File > Add Media, or click Add Media on the toolbar. In the Add Media dialog box, select all the.wav files to be converted and click Open to continue The Files to Convert window tab view appears.
Backup Master Files Before Converting
Be sure to make a backup copy of all master.wav files and store them in a separate directory away from any editing or converting activity. Do this to prevent losing master files by accidently converting the master files into the compressed.ivc or.wav format. The compressed files cannot be converted back to masters, due to quality loss during conversion.
Select Convert Files. The status changes to complete.
Modifying the Sample Batch Script
The Intervoice Example Batch Script file provides the Batch Converter with an example of the appropriate presets to be used during the conversion of master voice files into ADPCM (.ivc) or A-law or -law PCM.wav files. The batch script described below is available on the Intervoice Web site at www.intervoice.com/support/techinfo/voicerecording. This topic explains how to modify the sample batch script for different requirements. The basic script is designed to resample and compress.wav files into the runtime format. This sample script also includes settings to trim and normalize volume. The settings that can be modified are message level normalization and equalization. See page 5-9 for the settings that cannot be modified. To modify the batch script, follow these steps:
This is a general batch script explanation only. You need to change formats and volume levels for real runtime systems. For more information on batch scripts, refer to the Sound Forge document.
Open the Batch Converter. From the main Sonic Foundry Batch Converter window, select File > Open Script. Double-click the settings you want to change.
Rename and save the changed batch script.

Batch Script Settings That Cannot Be Modified
Some batch script settings should not be modified, as they control critical parameters in the conversion process. Other batch script parameters can be modified to adjust the runtime messages for various use. The settings that should not be modified are: DC offset removal Front and back silence trimming Front and back silence buffering 8 KHz resampling ADPCM compress for A-law playback Intervoice formatting
Filename Formats (.ivc to m.###), 6-2 Using the CRVoice2 Utility, 6-4
This chapter describes the Intervoice InterSoft IQTalk voice file naming convention and explains how to use the CRVoice2 utility to rename.ivc files to m.###. Most audio recording applications accept message files with long filenames such as BankABCgreeting503.wav. Messages can be given long filenames for easy identification. The Intervoice InterSoft IQTalk system only accepts messages in a specific format. The format is m.###. All InterSoft runtime messages must start with m, and end with three numeric digits. Message numbers larger than 999 are represented by placing more digits to the left of the suffix and to the right of the m. For example, message number 2046 is renamed to m2.048 and message number 48, 423 would be m48.423. The CRVoice2 utility can be used to quickly convert long filenames to the InterSoft runtime standard of m.###.
Filename Formats (.ivc to m.###)
Intervoice master files are.wav files. They have filenames of naming convention xxxxx.wav, where xxxxx can be alphabetic, numeric, alphanumeric, or other characters except special characters (for example, slashes and colons). The filenames have a limit of 253 characters, plus 3 character extension. Intervoice InterSoft IQTalk compressed files have two different naming conventions: xxxxx.ivc and m.###. Microsoft Windows programs identify the.ivc extension as the Intervoice file extension. The m.### naming convention is used to identify Intervoice messages placed on the Intervoice runtime system. However, Windows does not correctly identify files named m.###, so these files are not used on the Windows development or recording systems. The m.### files are used only on the Intervoice system. Figure 6-1 shows examples of accepted filename formats, including.wav,.ivc, m.###, and m#.### formats.

Original Master

BankABC502.wav

Firstmenu2046.wav

Compress to Intervoice Format

BankABC502.ivc

Firstmenu2046.ivc
Rename for Runtime (CRVoice2 Utility)
Rename for Runtime (3-character Extension)

m2.046

Figure 6-1. Filename Formats (.ivc to m.###)
Voice File Naming Convention

Before voice files can be used on an Intervoice system, the files must conform to Intervoices voice file naming conventions. InterSoft IQTalk runtime voice files must be named in the following format: m.###. The smallest message number allowed is m.000. Once m.999 is reached, begin renumbering at m1.000, m1.001, and so on. The highest message number supported in runtime is m2147483.647. Windows 2000 also requires all media files to have three-character extensions to identify the file type. Numeric file extensions are not recognized by Microsoft Windows, so Intervoice created a new extension,.ivc, to identify Intervoice files on Windows systems. However, when.ivc files are transferred to an Intervoice system, the.ivc extensions must be replaced with numeric extensions. The CRVoice2 utility renames the.ivc files to the m.### format. For Intervoice purposes, the.wav file extension indicates a master voice file, while the.ivc file extension indicates a file that has been compressed for use on an Intervoice InterSoft IQTalk system. Once you have compressed the messages into.ivc files, use the CRVoice2 utility (see page 6-4) to copy and rename all the.ivc files to the m.### format. The CRVoice2 utility leaves the original long filenames on the.wav and.ivc files for easy reference later. Figure 6-2 illustrates the progression from.wav files to.ivc files then to m.### files.
BankABC501.wav BankABC502.wav BankABC503.wav

Batch Converter

BankABC501.ivc BankABC502.ivc BankABC503.ivc

(original masters)

(trimmed, EQd, normalized, compressed) Same files, same content, different names CRVoice2 utility

m.501 m.502 m.503

Figure 6-2. Transforming.wav Files to.ivc Files
Using the CRVoice2 Utility
Before the.ivc voice files can be used on an Intervoice InterSoft IQTalk system, they must be renamed to the Intervoice voice filename format, m.###. The CRVoice2 utility automatically copies and renames all the.ivc compressed voice files created in Sound Forge to the Intervoice m.### file naming convention. CRVoice2 leaves the original long filenames on the.wav and.ivc files for easy reference later. The CRVoice2 utility also checks and corrects file lengths to ensure that all the files are exact multiples of four bytes. The CRVoice2 utility does not change the contents of a file in any way. If the.ivc filename includes the message number (for example, BankABCMainMenu503.ivc), the CRVoice2 utility copies and renames all message files quickly and correctly. This feature allows you to apply the full capabilities of long filenames to help identify message files before renaming them for placement on a runtime system. To use the default operation of the CRVoice2 utility, place the CRVoice2.exe file in the directory that contains the.ivc files to be renamed.

Frequently Asked Questions
This appendix provides some frequently asked message recording questions and the answers to those questions. Users should consult this chapter before calling for technical support. This chapter will be updated as new questions are encountered. Contact Sound Forge technical support at:
http://mediasoftware.sonypictures.com/support. (Intervoice recommends contacting Sonic Foundry first for any technical support issues, as Intervoice does not have specific support personnel for Sound Forge.)
For Intervoice technical support, see Technical Support on page xv. For a list of message recording FAQs, visit the Intervoice Web site at www.intervoice.com/support/techinfo/voicerecording.
The following questions and answers pertain to recording messages.

Recording Steps

What are the main steps to record a voice message?
The primary steps in the recording process are as follows:
Set levels. Perform the noise floor test (see page 2-12). Record messages (Refer to the voice recording software). Apply the batch script (see page 5-8). Run the CRVoice2 utility (see page 6-4), for InterSoft IQTalk only. Move the files to the Intervoice system.

6 Defining DC Offset

What is the DC offset and what does it do?
DC offset is the average vertical offset from 0 dB of the recorded wave form. This is normally visible in the Sound Forge window only in silent sections of the wave form. Most sound cards have some DC offset when you are recording. Usually you cannot see this offset in the Sound Forge window unless you expand the vertical gain nearly to the limit and look into a silent portion of the recorded wave form. DC offset can cause problems in compressing voice files using the Intervoice ADPCM format. It also causes errors when you are trying to measure the noise floor of a recording. The Sound Forge Tools > Statistics command measures the average value or DC offset of a wave form along with several other important statistics about a wave form. This is the command used to check the noise floor of your recording setup. Since virtually all sound cards have some DC offset, it should be routinely removed as part of the standard processing of 11.025 KHz master files for compression into Intervoice formatted messages. A preset to remove DC offset should be placed in the standard Intervoice batch script that is applied to all Sound Forge voice messages prepared for runtime systems.
Sound Forge already has a preset (Process/DC Offset) that removes the DC offset from any recording. Select Process > DC Offset in the Sound Forge main window. The DC Offset window opens. Ensure that the Automatically detect & remove option is selected and that the Compute from first 5 seconds only option is unchecked. Save this preset as Intervoice DC Removal Preset and place it in the standard batch script.

This is not an error or problem. The software is working correctly. Once the sound card is installed, the volume is controlled through the Midiman Audio Buddy preamplifier.

Turning Off AGC

How do I turn off the Automatic Gain Control (AGC) on my sound card?
The AGC that is on many sound cards must be turned off for any Intervoice voice message recordings. Unfortunately, Creative Labs turns on the Automatic Gain Control on all of their sound cards by default (except the AWE Gold). You can turn off AGC only by using the Creative mixer applet that comes with the Sound Blaster card in the application software package (ctmix32.exe). The standard Microsoft Windows 95 mixer applet (yellow speaker) does not have the ability to change this setting on Creative Labs cards. To turn off the AGC, follow these steps:
Install Creative Labs Audio Mixer applet (ctmix32.exe). Start ctmix32.exe. Right-click the mixer face. Select the Input/Output settings. Uncheck Automatic Gain Control.
You should not make recordings with the Sound Card AGC enabled. It can cause background noise as well as clicks and pops in the final recording.
Normalizing the Message Volume
When is the best time to normalize the volume so that it matches the messages already on the system?
The messages should be saved in 11, 025 KHz, 16-bit, monophonic, linear.wav format and recorded as loud as possible without clipping. All the subsequent processing can be performed using the Batch Converter. Generally, reducing the volume of the message is one of the last steps before compressing the.wav file into the runtime format.

Filtering Messages

When I am converting messages to the Intervoice ADPCM format, what filtering or optimization is necessary for the telephony network?
Converting the.wav Files to Runtime Format Overview on page 5-4 describes the steps for postprocessing a.wav file to InterSoft.ivc file format. The most critical step for voice clarity is equalization. By using the parametric equalization function in the Batch Converter to change the low frequency cutoff of the recorded voice, you can significantly change the clarity of the voice over the phone. Most users find that they can achieve their voice quality requirements by modifying the low-shelf frequency of the paragraphic EQ step. Usually they raise the low-shelf to between 200 and 300 Mhz for improved clarity. You may find that a modification of the high-shelf parameter also will improve some types of voices. If you want, you can boost a small portion of the audio spectrum centered at 800 Hz up a few decibels to get extra clarity voice. Do not improve the voice by boosting the bass. While this may seem improved on a hi-fi speaker, the voice quality does not work well over a phone handset. The primary information carrying bandwidth of the human voice is between 400 and 2000 Hz, and boosting frequencies outside of this range actually lowers the listeners comprehension.

A utility that truncates messages to a length divisible by 4 bytes. This truncation eliminates clicks and pops within messages used on the system. This utility is not recommended by Intervoice.

InterSoft system

An Intervoice system that operates on the InterSoft runtime environment with IQTalk or the VoiceXML Browser.

InterSoft version 3

The core software program used to bring applications into IQTalk voice browser runtime. InterSoft version 3 is designed to run on Windows NT and Windows 2000 operating systems. Intervoice IQ Environment [VoiceXML Browser System] A completely integrated environment that is bundled with the Intervoice InterSoft runtime system. However, it can also have speech recognition and text-to-speech servers. It offers a complete solution including design, development, and deployment tools as well as operation administration and management tools. It also includes data integration options to connect to legacy, enterprise server, and Web-based data sources. Browser options are IQTalk or VoiceXML, and operating system options are Windows NT or Windows 2000.
Intervoice runtime system
One or more Intervoice voice browser systems and other hardware that perform call processing, audiotex, speech recognition, call directing, voice messaging, and interactive voice response. This system is the interface between telephone callers and a host computer, the Internet, a local area network (LAN), a wide area network (WAN), or telecommunication network, and local or remote databases. The Intervoice voice browser systems can be integrated in one box or distributed in multiple boxes. This can also be called IVR and was formerly called Omvia Media Server.

IQTalk

The InterSoft runtime engine and voice browser that interprets and processes the InVision application. It can also interface with userdefined programs written in C or Assembly.
Kilobits per second. This is the number of bits passing a specific point per second in an audio file/voice message.

message

An audio file that, combined with other messages, comprises the spoken information in an application.

Glossary-3

Mu-law
See -law. A process used in the Intervoice Example Batch V2.sfz file to adjust the volume of all the audio files to the same level.

normalization

Omvia Media Server
See Intervoice runtime system.

preset

An audio processing function in Sound Forge.

RealCare

The remote diagnostic and troubleshooting service available to customers as a contractual option. RealCare operates a national WATS line at 1-800-955-4688.

Glossary-6

.ivc files 6-4, 6-5 naming 6-3 quality of 7-2.wav files 6-4 clipping 2-2 conversion and reconversion 7-2 converting 5-2 defined Glossary-5 described 3-2 editing and compressing 7-2 master format 7-2 keeping two copies 7-2 renaming 6-3 updating 7-2 -law compressed data format, explained 3-2 A AC-powered speakers with cables, recommended hardware 2-3 adaptive pulse code modulation (ADPCM) 5-2.ivc data format 3-3, 5-8 decoder 5-4 ADPCM. see adaptive differential pulse code modulation A-law ADPCM compression 5-5 defined Glossary-1 applet, ctmix32.exe A-5 audience for this document ix audio cable recommended hardware 2-3 Y-adapter 2-4
audio file format InterSoft IQTalk 3-3 VoiceXML 3-3 Automatic Gain Control (ACG) A-5 automatic labeling and Sound Forge 4-2 defined Glossary-1 setting up 4-2 B background noise 2-12 bandwidth, defined Glossary-1 bass boosting A-6 Batch Converter 3-3, 4-6 defined Glossary-2 downloading 2-6 installing 5-2 modifying sample batch script 5-8 parametric equalization function A-6 postprocessing steps 8 KHz resampling 5-4 DC Offset Removal 5-4 Front and Back silence buffering 5-4 Front and Back silence trimming 5-4 Message level normalization 5-4 registering 2-6 removing batch scripts 4-18 using 5-6 using sample batch script 5-3 batch scripts modifying 5-3, 5-8 removing 4-18 buffering, silence 5-4

Index-1

buttons Record 2-8, 4-2 Reset 2-8 Stop 4-8 C caution note xiv checklists recording process 4-6 studio set up 2-2 click defined Glossary-2 reducing A-6 clipping described 2-9 fixing 2-9 preventing 2-10 command prompt 6-4 defined Glossary-2 connecting the sound card and speakers 2-4 conventions used in this document xiv converting.wav file to.ivc file 5-4 CRVoice2 utility 4-6 and.ivc files 2-7 command 6-5 defined Glossary-2 install directory 6-4 options <-rename> (optional) 6-5 <source file spec> 6-5 <target directory> 6-5 -h 6-5 renaming files 2-7 starting from a command prompt 6-5 starting from an icon 6-5 using 6-4 CRVoice2.exe file 6-4, 6-5 ctmix32.exe A-5 customer support xv
D dB(decibel), defined Glossary-2 dBm, defined Glossary-2 DC offset defined Glossary-2 removal 2-13, 5-4 statistics, viewing 2-13 window 2-12 diagram conversion process 5-2 creating audio recordings 1-3 from.wav to.ivc 6-3 DMA (direct memory access), defined

document reference xiv documentation access xi formats xi Intervoice xi OEM for installation instructions 2-2 downloading from Web site Batch Converter 2-6 E equalization 5-5 defined Glossary-2 example Batch Script 5-2 F FAQs converting messages A-3 defining DC offset A-2 filtering messages A-6 installing the Audiophile 2496 card A-4 normalizing the message volume A-5 producing 64-KB messages A-3 recording steps A-2 reducing pops and clicks A-6 turning off Automatic Gain Control (AGC)
using the Gain Leveling Preset A-4 figures, list of v filename defined Glossary-2 formats 6-2

Index-2

files.ivc 6-5.wav defined Glossary-5 CRVoice2.exe 6-4, 6-5 Intervoice Example Batch Script 5.0 5-3 filtering messages A-6 formats, standard InterSoft runtime 3-3 G Gain Leveling Preset, described A-4 general note xiv H hardware, recommended 1-4 I information symbols used in this document xiv installing Batch Converter 2-6 CRVoice2 2-7 hardware 2-3 InterSoft file naming conventions 6-1 master format, differences between Windows master format 3-3 runtime format 3-2 runtime messages, format 6-1 InterSoft system, defined Glossary-3 InterSoft version 3, defined Glossary-3 Intervoice documentation xi Example Batch Script 5.0 file 5-3 Example Batch Script file contents 5-3 Example Batch V2.sfz file 5-2 master format 3-2 standard master format 3-2 technical support xv voice file, naming conventions 6-3 Web site 1-1, 5-3, 5-8, 6-5, A-1 Intervoice system defined Glossary-3 IQTalk, defined Glossary-3
J jack-to-jack shield adapter, recommended hardware 2-3 K KBps defined Glossary-3 kilobits per second Glossary-3 L list of figures v list of tables vii M master format.wav files 7-2 Windows 3-3 M-audio sound card, recommended 2-3 message defined Glossary-3 file, final volume levels 2-8 filtering A-6 volume, normalizing A-5 microphone 2-2 cable, recommended hardware 2-3 commercial and preamplifiers 2-9 connecting 2-2 level on preamplifier 2-9, 2-12 placement 2-9 recommended hardware 2-3 recommended type, supercardioid 2-9 Sennheiser, more information 2-9 setting levels 2-8 modifying the example batch script 5-8 mu law, defined Glossary-1 N noise floor test 2-2, 2-12, 4-6 failure causes 2-13 normalization 5-4 defined Glossary-4 Normalize preset A-4

Index-3

note boxes caution xiv document reference xiv general xiv tip xiv warning xiv O Omvia Media Server defined Glossary-4 optimization of messages A-6 P parametric equalization function A-6 playback 5-5 plug-in chain 5-4 pop defined Glossary-4 reducing A-6 postprocessing 5-4 power cord, recommended hardware 2-3 preamplifier 2-2 and commercial microphones 2-9 line-level input aux 2-2 line 2-2 recommended hardware 2-3 preset defined Glossary-4 Gain Leveling A-4 Normalize A-4 prompt, command, defined Glossary-2 purpose of this document ix R RealCare xv defined Glossary-4 recommended hardware AC-powered speakers with cables 2-3, 2-4 audio cable 2-3 jack-to-jack shield adapter 2-3, 2-4 M-audio sound card 2-3 microphone 2-3 microphone cable 2-3 power cord 2-3 preamplifier 2-3

Sound Forge on InterSoft Systems Users Guide

60000984First Edition

Document No. 60000984 First Edition
Copyright 2001 InterVoice-Brite, Inc. All rights reserved. Trademarks of InterVoice-Brite, Inc. and/or its affiliates:
Names and marks for products provided by the Companys suppliers are used herein for identification purposes and may be trademarks of their respective companies. Windows NT and Windows 2000 are registered trademarks of Microsoft. The Apache Software License, Version 1.1 Copyright 20002001 The Apache Software Foundation. All rights reserved. Java Runtime Environment Versions 1.2 and 1.3 are products of Sun Microsystems, Inc. All rights reserved. Please refer to the license agreement on your system. No part of this manual may be reproduced without the written permission of InterVoice-Brite, Inc.
The information in this manual explains how to create audio files for InterSoft versions 2 and 3 systems. It includes Sound Forge 5.0 and CRVoice2. If earlier software versions are used, refer to Using Sound Forge for InterVoice Systems (document number 60000259) released in June 1999. An archived copy is available on the InterVoice-Brite Web site. This is the first edition of this manual, December 2001.
Any comments or suggestions concerning this publication should be directed to: Technical Documentation Department InterVoice-Brite, Inc. 17811 Waterview Parkway Dallas, Texas 75252 Telephone: (972) 454-8200 Fax: (972) 454-8905 E-mail: document.feedback@intervoice-brite.com For more information about InterVoice-Brite, Inc., call (972) 454-8000, or visit the InterVoice-Brite Web site at www.intervoice-brite.com.
Documentation Comment Form
Our goal is to provide accurate and efficient information to our customers. Please complete this form to help us improve the quality and usability of this product documentation. List the document you used by name, number, and edition.
____________________________________________________________________ ____________________________________________________________________ Did the document provide the information needed?

Yes No

Explain______________________________________________________________ ____________________________________________________________________ ____________________________________________________________________ Were you able to locate the information in a timely manner?
Explain______________________________________________________________ ____________________________________________________________________ ____________________________________________________________________ How accurate did you find the information in the document to be?
Very Accurate Accurate Somewhat Accurate Not Accurate Table of Contents Index Headings Scan the text Yes No
When you need to find information, where is the first place you usually look?

Was the index thorough?

How could it be improved?

____________________________________________________________________ ____________________________________________________________________ If you completed the steps in this manual, how easy were they to follow?
Very Easy Easy Somewhat Easy Not Easy
Overall, how helpful did you find the document to be?
Very Helpful Helpful Somewhat Helpful Not Helpful
Are there any technical errors in the document you would like to report?
____________________________________________________________________ ____________________________________________________________________ List any additional topics you would like to see covered in the documentation.
____________________________________________________________________ ____________________________________________________________________ Fax to: (972) 454-8905 E-mail to: document.feedback@intervoice-brite.com

Table of Contents

List of Figures List of Tables About This Manual

v vii ix

Purpose and Audience... ix Chapters... x InterVoice-Brite Documentation.. xii Related Manuals... xiii InterSoft Version 3... xiii InterSoft Version 2... xiv Sonic Foundry... xv Conventions.... xvi Information Symbols.. xvi Technical Support.... xvii

Overview 1-1

Voice Recording Concept Diagram... 1-2 Hardware Requirements... 1-3 Specific Hardware.. 1-3 Software Requirements... 1-4
Setting Up the Studio 2-1
Studio Setup Checklist... 2-2 Installing the Hardware... 2-3 Connecting the Sound Card to the Speakers.. 2-4 Installing the Sonic Foundry Batch Converter.. 2-6 Installing CRVoice2... 2-6 Setting Microphone Levels... 2-7 Microphone Placement... 2-7 Clipping.... 2-7 Microphone Levels... 2-9 Performing the Noise Floor Test.. 2-10
Defining Audio File Formats
InterVoice-Brite Master Format.. 3-2
Naming and Converting Files 4-1
Filename Formats... 4-2 Voice File Naming Convention.. 4-3 Using the CRVoice2 Utility... 4-4
Maintaining Audio Quality 5-1
Voice File Quality Issues..5-2

Postprocessing 6-1

Using the Batch Converter... 6-2 Installing the Sample Batch Script.. 6-2 Converting the.wav Files to.ivc Format.. 6-3 Using the Batch Converter... 6-5 Modifying the Sample Batch Script... 6-7

Chapter 1Overview

Provides the overview for this manual. It contains a purpose statement for the entire manual. It also contains an audio recordings concept diagram and hardware requirements.
Chapter 2Setting Up the Studio
Describes the elements necessary to properly set up the recording studio. It explains how to install the voice recording hardware and configure volume controls. It also discusses how to set recording attributes and the microphone level, and how to perform the Noise Floor Test.
Chapter 3Defining Audio File Formats
Describes the voice file formats and explains the master format differences.
Chapter 4Naming and Converting Files
Describes the InterVoice-Brite voice file naming convention and explains how to use the CRVoice2 utility.
Chapter 5Maintaining Audio Quality
Explains the purpose of master voice files and stresses the importance of editing the files before compressing them. It also stresses the importance of retaining original master voice files.

Chapter 6Postprocessing

Provides the steps to convert multiple voice files to the voice file format using the Sonic Foundry Batch Converter 5.0.
Chapter 7Using Sound Forge to Record Messages
Describes how to set up the automatic labeling functions and describes the voice recording process, the methods of recording voice messages, and the use of Sound Forge to record voice messages.
Appendix AFrequently Asked Questions
Provides some frequently asked message recording questions and the answers to those questions. Users should consult this chapter before calling for technical support. This chapter will be updated as new questions are encountered.
InterVoice-Brite Documentation
InterVoice-Brite provides its standard product documentation on CD-ROM. The manuals on the CD-ROM are in Adobe Acrobat format (.pdf files) and can be viewed, navigated, and printed. The CD-ROM includes: Adobe Acrobat Reader with the Search tool plug-in, which enables viewing and full-text searches of the documents. You can view the document files using the copy of Acrobat Reader installed on the CD-ROM, or you can install Acrobat Reader on your local hard drive. A readme.txt file that lists system requirements and explains how to install Acrobat Reader. A welcome file that provides navigation to the document files.
For more information about using Adobe Acrobat Reader, see the welcome file on the CD-ROM, or go to the Adobe Web site at http://www.adobe.com.
The following core product CD-ROMs are currently available:

Core Software Documentation CD-ROM Doc Number 62000047
Provides a complete, easy-to-use reference of the core software product technical documents, including core features, CTI interface, database interface, host interface, software installation, InterSoft, InVision, network connectivity, history reporting, system management, and telephony product manuals.
Core Installation Documentation CD-ROM Doc Number 62000083
Provides a complete, easy-to-use reference of the core installation product technical documents, including site preparation, installation, maintenance, system software, system hardware, and system configuration manuals. The following technical documentation catalog is shipped along with the CD-ROM:
Core Product Documentation Guide Doc Number 60000021
Describes the documents available with each product line and explains how to access them on a technical documentation CD-ROM. InterVoice-Brite also provides an online source of its documentation through the company Web site. RealCare support customers, customers under warranty, resellers, and distributors can visit http://www.intervoice-brite.com/support to register for access to the documentation. Hard copy manuals or additional CD-ROMs can be purchased from any InterVoice-Brite sales representative.

Related Manuals

The following manuals contain useful information about your product. The manuals are listed in functional order.

InterSoft Version 3

InterSoft System Administration (InterSoft version 3) Doc Number 60001092
Explains how to prepare an InterSoft system for the runtime production environment. It also provides commands and explanations of the basic system administration functions of an InterSoft version 3 system, including starting and stopping the runtime system, monitoring system status, creating and editing the various application tables, and working with voice messages.
InterSoft System Utilities (InterSoft version 3) Doc Number 60001094
Describes the system utilities used to prepare, operate, and monitor a system in runtime.
Monitoring Runtime Status (InterSoft version 3) Doc Number 60001093
Describes the tools available for monitoring, testing, and debugging a runtime system. These tools include the various IQTalk windows, the IBROWSE utility, and the IvIViews tool.

Note Batch Converter Update 5.0a must be installed before
installing and using the IVB Example Batch Script file. For more information, contact Sonic Foundry technical support at (608) 256-5555, fax: (608) 256-7300, e-mail: support@sonicfoundry.com, or visit the Sonic Foundry Web site at www.sonicfoundry.com.

Installing CRVoice2

Numerical file extensions are not recognized correctly by Windows, so InterVoice-Brite created a proprietary extension,.ivc, to help identify InterVoice-Brite files on Windows systems. However, when.ivc files are transferred to an InterSoft system, the.ivc extensions (for example BankABC502.ivc) must be replaced with numeric extensions such as m.502. The CRVoice2 utility renames the.ivc files to m.### files. The CRVoice2 utility is available on the InterVoice-Brite Web site at www.intervoice-brite.com/support/techinfo/voicerecording.
Setting Microphone Levels
To ensure that the microphone levels are set correctly, the microphone must be properly placed and clipping must be avoided. As a general rule, microphone levels should be as loud as possible without causing any clipping. You can adjust final volume levels in message files using the volume or dynamic graphic preset after all messages are recorded. The best signal is the noise ratio are achieved when a loud, unclipped recording is made.

Microphone Placement

InterVoice-Brite recommends a supercardioid vocal microphone (such as the Sennheiser E845) to help eliminate background noise during recording. With a desk-mounted microphone, ensure that the distance between the voice talents mouth and microphone is fairly constant for each recording. Most commercial microphones do not have an output signal level that is high enough to drive the sound card and, therefore, require a microphone preamplifier. External microphone preamplifiers improve signal-to-noise ratios by keeping high-gain amplification stages out of the PCs electrical noise fields. For more information about the Sennheiser E845, go to www.allprosound.com or www.fullcompass.com.

10 Verify that the Average value (DC offset) statistic is 0.
Perform this test regularly on systems used to record voice messages. The voice message fails the test if the noise floor is louder than 60 dB. Louder noise is represented by lower negative number.
For example, 55dB noise floor is louder than 60 dB. The test may fail because the room noise is too loud (due to air conditioning, PC fan, paging system), or the sound card preamplifier is noisy. Correct any problems before continuing to record. Noise floors greater than 60 dB (for example, 55 or 50 dB) can be heard in the playback process and cause pops and other audible degradation during the ADPCM compression process.
InterVoice-Brite Master Format, 3-2
This chapter describes the voice file formats and explains the master format differences. The voice files are in.wav and.ivc formats.
InterVoice-Brite Master Format
InterVoice-Brite uses the.wav format for recording and saving all voice files. A.wav file is a standard format in which messages are recorded on a computer. Almost any audio recording tool designed for computers can be used to create the.wav file format. The InterVoice-Brite standard master format uses the.wav file format because this format supports high-quality recordings, and the equipment required to record and play.wav files is commonly available. The InterVoice-Brite master.wav format is 11,025 samples per second, 16 bits per sample, linear encoding, which equates to a data rate of 22,000 bytes per second. The InterVoice-Brite master format keeps the master files in a much higher quality format than the older compressed -law masters. This format does make the file sizes larger, but the voice quality of subsequent copies of this master is much better.
InterVoice-Brite Standard Master Format
The InterVoice-Brite standard format for recording master voice files is 11,025 kHz, 16-bit, monophonic, linear.wav file. This format is the master format for storage of InterVoice-Brite voice files.

Filename Formats

InterVoice-Brites master files are.wav files and have filenames of xxxxx.wav, where xxxxx can be any characters except special characters (for example, slashes and colons) up to 253 characters. InterVoice-Brite compressed files have two different naming conventions: xxxxx.ivc and m.###. Microsoft Windows programs identify the.ivc extension as the InterVoice-Brite file extension. The m.### naming convention is used to identify InterVoice-Brite messages placed on the runtime system. However, Windows does not correctly identify files named m.### , so these files are not used on the Windows development or recording systems. The m.### files are used only on the InterSoft runtime system. Figure 4-1 shows examples of accepted filename formats, including.wav,.ivc, m.###, and m#.### formats.

Original Master

BankABC502.wav

Firstmenu2046.wav

Compress to InterVoice-Brite Format

BankABC502.ivc

Firstmenu2046.ivc
Rename for Runtime (CRVoice2 Utility)
Rename for Runtime (3-character Extension)

m2.046

Figure 4-1. Filename Formats
Voice File Naming Convention
Before voice files can be used on an InterSoft runtime system, the files must conform to InterVoice-Brites voice file naming conventions. InterSoft runtime voice files must be named in the following format: m.###. The smallest message number allowed is m.000. Once m.999 is reached, begin numbering at m1.000, m1.001, and so on. The highest message number supported in runtime is m2147483.647. Windows NT and Windows 2000 also require all media files to have three-character extensions to identify the file type. Numeric file extensions are not recognized by Microsoft Windows, so InterVoice-Brite created a new extension,.ivc, to identify InterVoice-Brite files on Windows systems. However, when.ivc files are transferred to an InterSoft runtime system, the.ivc extensions must be replaced with numeric extensions. The CRVoice2 utility renames the.ivc files to the m.### format. For InterVoice-Brite purposes, the.wav file extension indicates a master voice file, while the.ivc file extension indicates a file that has been compressed for use on an InterSoft runtime system. Once you have compressed the messages into.ivc files, use the CRVoice2 utility (see page 4-4) to copy and rename all the.ivc files to the m.### format. The CRVoice2 utility leaves the original long filenames on the.wav and.ivc files for easy reference later. Figure 4-2 illustrates the progression from.wav files to.ivc files then to m.### files.

The InterVoice-Brite Example Batch Script 5.0 creates 32-KB compressed messages for -law playback and 12 dB volume normalization. You may obtain other options, such as -law playback, different volumes, and 24-KB compression, by editing the script in the Sonic Foundry Batch Converter.
Converting the.wav Files to.ivc Format
A series of transformational steps is required to convert a.wav file into the InterSoft runtime.ivc file format. All of these steps can be automated into a single batch process in the Batch Converter. The Batch Converter can apply this process to a large number of master.wav files in a single step, creating a set of InterSoft runtime.ivc files, one for each master.wav file. To convert a master.wav file into an InterSoft.ivc runtime file, the Batch Converter must perform a specific set of processes in a specific order on the original.wav master file. In version 5.0, Sonic Foundry calls these steps plug-ins (presets in Sound Forge 4.5). The final script is called a plug-in chain. The basic recording process provides a clean audio signal digitized to the systems hard drive in.wav file format. However, there are several other processing steps that must be applied to an audio file after it is recorded but before it is suitable to be placed on an InterSoft runtime system. The InterVoice-Brite sample Batch Converter postprocessing procedures are described below:

DC Offset removal

Most sound cards have some DC offset that can affect the compression process. This step removes any DC offset before compressing messages.
Front and Back silence trimming with 20-ms fades
This step removes excess silence on the front and back of messages to tighten message delivery. Generally the messages are trimmed to 20 ms or less of silence on both ends. Noise at the beginning or end of messages (such as mouse or keyboard clicks) causes the autotrimming process to trim too early or too late, leaving extra silence in the message. These messages must be hand-trimmed.
Front and Back silence buffering with 50-ms buffers
Once the arbitrary silences are trimmed from the message, this step places a standard 50-ms silence buffer on the beginning and end of every message. This prevents the destabilization of the ADPCM decoder when two messages are played back to back.
Message level normalization
This step converts all the messages in a single application to a single standard volume level, so that listeners hear every message at the same volume.

8-kHz resampling

This step converts the 11-kHz sample rate of the master file to an 8-kHz sampling rate to prepare for ADPCM conversion.

Equalization

This step performs the frequency manipulation required to make high-quality message playback over telephone lines.
ADPCM compression for A-law playback and InterVoice-Brite formatting

This step compresses the linear format to ADPCM to reduce the size of the message files on the InterSoft runtime system. This is where you specify whether a message is to be played as -law or A-law, or to be compressed to 24 or 32 KB.
To use the Batch Converter, follow these steps:
From the main Sonic Foundry Batch Converter window, select File > Open. The Sonic Foundry Batch Converter - IVB Example Batch Script V5.0.bcs window appears.
Do not select any of the options on the window. Select the directory where you want the output files saved. Select the file format template and description.
Note If you change any selection, save the changes as a
new batch script and retain the original sample batch script.
Select the files that are to be converted from the main Sonic Foundry Batch Converter window. Select File > Add Media. The Add Media window appears.
Select all the.wav files that you want to convert. Click Open to continue.
Select Convert Files. The status changes from new to complete.
Modifying the Sample Batch Script
The InterVoice-Brite Example Batch Script 5.0 file provides the Batch Converter with an example of the appropriate presets to be used during the conversion of master voice files to InterVoice-Brite ADPCM (.ivc) files. The batch script described below is available on the InterVoiceBrite Web site at www.intervoicebrite.com/support/techinfo/voicerecording. This topic explains how to modify the sample batch script for different requirements. The basic script is designed to resample and compress.wav files into the.ivc runtime format. This sample script also includes settings to trim and normalize volume. The settings that can be modified are message level normalization and equalization. See page 6-8 for the settings that cannot be modified. To modify the batch script, follow these steps:
This is a general batch script explanation only. You need to change formats and volume levels for real runtime systems. For more information on batch scripts, refer to the Sound Forge 5.0 Manual.
Open the Batch Converter. From the main Sonic Foundry Batch Converter window, select File

> Open.

Double-click the settings you want to change.
Rename and save the changed batch script.
Batch Script Settings That Cannot Be Modified
Some batch script settings should not be modified, as they control critical parameters in the conversion process. Other batch script parameters can be modified to adjust the InterSoft runtime messages for various use. The settings that should not be modified are as follows: DC offset removal Front and back silence trimming Front and back silence buffering 8 kHz resampling ADPCM compress for A-law playback InterVoice-Brite formatting
Installing the InterVoice-Brite Batch Script File, 7-2 Setting Up Automatic Labeling, 7-6 Setting Recording Attributes, 7-8 Recording Process Checklist, 7-9 Recording Methods, 7-10 Recording Messages, 7-12 Reviewing Messages, 7-15 Removing InterVoice-Brite Batch Scripts, 7-16

Create a New Window for Each Take

Recording Messages

creating Regions
To record master voice files in Sound Forge using the Multiple takes mode, follow these steps:
Set up the Automatic Labeling function (see page 7-6). If you are using Sound Forge version 4.5 or 5.0, make sure you have set up the Automatic Labeling function. If you skip this step, you will have to manually number each of your regions or takes after recording.
Open Sound Forge. Open the Record window by either clicking the Record button (the red dot) or choosing Special > Record. The Record window appears.

Input level

Monitor

Record

In the Mode field, select the recording mode.
Note Make sure that DC adjust, in the lower right of the
Record window, is NOT checked.

Regions

If you have the Batch Converter plug-in, Multiple takes creating is the easiest recording mode to use. This mode places all takes in a single window, marking the beginning and end of each new take with a region marker and a vertical black line. Later, using the Extract Regions tool (available only with the Batch Converter plug-in), automatically saves each region as a separate file. When recording large numbers of messages (over 50), you may want to divide the regions into several different files for easier manipulation. However, if you do not have the Batch Converter plug-in, use New mode. This mode places each take in its own data window. To view all your takes, choose Window > Cascade > Tile Horizontally, or Window > Cascade > Tile Vertically.

window for each take

Select the Monitor checkbox. This turns on the input monitor which displays the level of the recording input on the meters shown in the window on page 7-12. When a recording is clipped, the word Clip appears above each meter. Clipping destroys a recordings quality. The Monitor function allows you to set the appropriate volume at which to record to prevent clipping.
Set the recording volume using either the sound cards or the operating systems volume control application. The volume should be set at the highest level at which you can speak at normal volume without clipping. If you are recording in Create a new window for each take mode, click the Prepare button. This function allows time for you to create and rename the new window before Sound Forge begins recording. If you do not use the Prepare button, the beginning of each message may be cut off. It is not necessary to use this function if you are using the Multiple takes creating Regions mode. After you click Prepare, a flashing green box that reads Prepare appears to the left of the Sync button.

The following questions and answers pertain to recording messages.

Note Recording Steps

What are the main steps when recording a voice message?
The primary steps in the recording process are as follows:

Defining DC Offset

Set levels. Perform the noise floor test (see page 2-10). Record messages (Refer to the voice recording software). Apply the batch script (see page 6-2). Run the CRVoice2 utility (see page 4-4). Move the files to the InterSoft runtime system.
What is the DC offset and what does it do?
DC offset is the average vertical offset from 0 dB of the recorded wave form. This is normally visible in the Sound Forge window only in silent sections of the wave form. Most sound cards have some DC offset when you are recording. Usually you cannot see this offset in the Sound Forge window unless you expand the vertical gain nearly to the limit and look into a silent portion of the recorded wave form. DC offset can cause problems in compressing voice files using the InterVoice-Brite ADPCM format. It also causes errors when you are trying to measure the noise floor of a recording. The Sound Forge Tools > Statistics command measures the average value or DC offset of a wave form along with several other important statistics about a wave form. This is the command used to check the noise floor of your recording setup. Since virtually all sound cards have some DC offset, it should be routinely removed as part of the standard processing of 11-kHz master files for compression into InterVoice-Brite formatted messages. A preset to remove DC offset should be placed in the standard InterVoice-Brite batch script that is applied to all Sound Forge voice messages prepared for runtime systems.
Sound Forge already has a preset (Process/DC Offset) that removes the DC offset from any recording. Select Process > DC Offset in the Sound Forge main window. The DC Offset window opens. Ensure that the Automatically detect & remove option is selected and that the Compute from first 5 seconds only option is unchecked. Save this preset as InterVoice-Brite DC Removal Preset and place it in the standard batch script.

Producing 64-KB Messages

Is it possible to produce 64-KB -law or A-law messages with the Sound Forge Tool?
Yes, you can create 64-KB messages with Sound Forge, as follows: Before you can save a Sound Forge file as an InterVoice-Brite 64-KB message, you must make sure the file is sampled at exactly 8.0 kHz. If the file is not sampled at 8 kHz, you must resample the message in Sound Forge to change the sample rate to 8 kHz. Remember that if you are down-sampling (for example, 22 kHz to 8 kHz), you must use the antialias filter in the resample preset. Upsampling (for example, 6 kHz to 8 kHz) does not require the antialias filter. (See Using the Batch Converter on page 6-2 for more information.) Once you have an 8-kHz message, select File > Save as. The Save As dialog box appears. In the Save as type field, select the InterVoice (*.ivc) format. In the Format selection field, select the G.711 -law or A-law format. Click the Save button. You can also put all of this into a batch converter script to convert several.wav voice files to InterVoice-Brite 64-KB master files in a single step.

Install Creative Labs Audio Mixer applet (ctmix32.exe). Start ctmix32.exe. Right-click the mixer face. Select the Input/Output settings. Uncheck Automatic Gain Control.
You should not make recordings with the Sound Card AGC enabled. It can cause background noise as well as clicks and pops in the final recording.
Normalizing the Message Volume
When is the best time to normalize the volume so that it matches the messages already on the system?
The messages should be saved in 11, 025 kHz, 16-bit, monophonic, linear.wav format and recorded as loud as possible without clipping. All the subsequent processing can be performed using the Batch Converter. Generally, reducing the volume of the message is one of the last steps before compressing the.wav file into the InterSoft runtime format.

Filtering Messages

When I am converting messages to the InterVoice-Brite ADPCM format, what filtering or optimization is necessary for the telephony network?
Converting the.wav Files to.ivc Format on page 6-3 describes the steps for postprocessing a.wav file to InterSoft.ivc file format. The most critical step for voice clarity is equalization. By using the parametric equalization function in the Batch Converter to change the low frequency cutoff of the recorded voice, you can significantly change the clarity of the voice over the phone. Most customers find that they can achieve their voice quality requirements by modifying the low-shelf frequency of the paragraphic EQ step. Usually they raise the low-shelf to between 200 and 300 Mhz for improved clarity. You may find that a modification of the high-shelf parameter also will improve some types of voices. If you want, you can get really fancy and boost a small portion of the audio spectrum centered at 800 Hz a few decibels to get extra clarity voice. One of the mistakes is to improve the voice by boosting the bass. While this may seem improved on a hi-fi speaker, the voice quality does not work well over a phone handset. The primary information carrying bandwidth of the human voice is between 400 and 2000 Hz, and boosting frequencies outside of this range actually lowers the listeners comprehension.
Reducing Pops and Clicks in Messages
I have heard that the CRVoice2 utility fixes the pops and clicks in messages. Is this a result of making sure the message is in 4-byte increments, or does something else resolve this problem? And what does the CRVoice2 utility do?
Sound Forge version 4.5 and later versions of the Batch Converter with the IVB Example Batch Script V5.0.bcs file create InterSoft.ivc files with the correct byte count lengths. If you convert your.wav message masters using Sound Forge 4.5 or later versions of the Batch Converter, which is the correct way to convert messages, you do not have problems with pops caused by message length. Other wave editors do not process or compress the message files correctly and should not be used to create InterSoft.ivc files. Other editors can be used to record the.wav master files, but only the Sonic Foundry Batch Converter can convert them.

The CRVoice2 utility checks the message length, and corrects it when necessary. However, it should not be used for that purpose because it does not perform the other operations on the.wav file that the Batch Converter does. The CRVoice2 utility is used strictly to rename the.ivc files to m.### for InterSoft runtime systems. That is all it is designed to do. Checking the message length was included in this utility to fix a bug in an earlier version of Sound Forge. This earlier version is no longer available.

Glossary

-law (pronounced mu-law)
A Pulse Code Modulation (PCM) voice-encoding standard in which analog voice is band-limited to 4-kHz bandwidth and then sampled at twice its highest frequency (4000 x 2) with the sample then converted as an eight-bit binary number. Samples must be taken 8000 times per second so each sample is 1/8000 of a second, or 125 microseconds. A standard telephone voice channel is transmitted at 64 kbps, or eight samples in 125 microseconds. The -law standard is used in the United States, including North America (on T1 systems) and in Japan, Mexico, and Canada. See also A-law.
adaptive pulse code modulation (ADPCM)
A method of encoding digitized voice that requires fewer bits than linear or -law encoding (voice compression). The InterSoft system has a standard file format for voice messages using this compression method at 32 kbps and 24 kbps per second.
The method of encoding sampled audio waveforms used in the 2048-KB, 30-channel PCM primary system. A-law is the standard widely used outside of North America on E1 systems. See also -law.

automatic labeling

The function that allows the user to assign descriptive filenames and set up automatic increments when message numbering master voice files. Using the Options menu, this function can be defined before recording messages.

bandwidth

The range of electrical frequencies a device can employ.
The Sonic Foundry program that allows the processing of multiple audio files in a single process.

Glossary-1

A sound, or noise, that occurs in voice recordings; a type of audible interference that must be corrected in voice recordings.

CRVoice2 utility

A utility that assists in the naming of multiple voice files for use in InterSoft runtime.
A decibel below 1 milliwatt (mW).

DC offset

The average of all of the positive and negative integers in a waveform. This number can be positive or negative. Zero is referred to as no DC offset.