Voice Data Compression; Automatic Detection Of Voice, Fax, And Data - ZyXEL Communications U-1496 series User Manual

computer stores and forwards. It also means the modem can playback the recorded digitized
voice either off-line for local message listening or on-line for a message announcement. For
interactive voice applications, DTMF tone detection capability is important so a computer can
react according to the remote caller's touch tone input.

VOICE DATA COMPRESSION

The main issue in digitized voice is the amount of storage required. A good phone quality voice
digitization will produce about 64 Kbits of data for each second of voice. The hard disk will
quickly be filled up by digitized voice at this digitization rate. Speech compression is needed to
reduce the digitized data rate. A relatively simple ADPCM (Adaptive Differential Pulse Code
Modulation) algorithm can reduce the speech data rate to half the rate and maintain about the
same voice quality. This algorithm can also be used to reduce the speech data rate to 1/3 or 1/4
of the original rate, but with voice quality degradation. Reducing the speech data rate further and
maintaining good voice quality requires a sophisticated and complicated signal processing
algorithm. It also requires a lot of digital signal processing computation power. We call this
sophisticated speech data compression capability advanced voice capability.
The U-1496 series modems support three voice digitization schemes. Two schemes use the
ADPCM algorithm and one uses the advanced CELP (Code Excited Linear Prediction) algorithm
to achieve near-phone-quality voice at a 9.6 kbps speech data rate. A summary of these three
schemes is listed below:
Digitization Speech Compression
Scheme Algorithm
CELP Code Excited Linear Prediction 9600 bps
2-ADPCM ADPCM, 2 bits/sample
3-ADPCM ADPCM, 3 bits/sample
where the sampling rate used is 9600 samples per second.
The ADPCM voice modes are supported on all U-1496 series modems. However, the advanced
CELP scheme is supported only on the plus models, including the U-1496 LCD model.
Silence detection is implemented to eliminate voice data coding in silent periods.
The modem also supports simultaneous DTMF (Dual Tone Multiple Frequency) tone detection,
thus enabling the use of the keypad of a tone-dialing push-button telephone set to control,
instruct, or answer the modem.

AUTOMATIC DETECTION OF VOICE, FAX, AND DATA

Since there is no standard and there is no way to specify how a human voice should behave in
telephone calling and answering, it is difficult, if not impossible, to automatically make a
distinction between voice calls, fax, or data calls. A common method to do this detection and
switch is the use of a human voice announcement as the first step in the answer process. The
calling party is instructed to push one number in the push-button pad for fax calling and push
another number for voice calling. By detecting which DTMF tone is received, the computer
software and the modem can switch to the appropriate mode accordingly. If no tone is received
until timeout, the modem assumes the call to be a data call and starts modem handshaking. The
shortcoming of this method is that some modems will get confused by the initial voice message
Data Rate
19200 bps
28800 bps