VoIP Voice and Fax Signal Processing

3.5 WIDEBAND VOICE

VoIP wideband voice is possible if an end-to-end call is only VoIP and not routed through PSTN. It has to be a direct VoIP call between wideband voice interfaces, including the acoustics. In wideband voice, acoustic interfaces and processing the signal can support frequencies from 50 to 7000 Hz sampled at 16 kHz. Traditional PSTN and VoIP service limits the analog speech bandwidth to 300 to 3400 Hz. Lower frequency enhancement (50–300 Hz) provides greater naturalness, presence, and comfortable listening. High-frequency enhancement (3400 to 7000 Hz) provides better intelligibility and fricative differentiation (e.g., s versus f). The implementation aspects of a wideband VoIP voice call are given in Chapter 9.

3.5.1 G.722 Codec

G.722 [ITU-T-G.722 (1988)] is a time-domain wideband-waveform-based codec that can work on a minimum of a block of two samples. In actual implementations, a block of 5 or 10 ms (160 samples) is used for processing. G.722 uses sub-band adaptive differential pulse code modulation (SB-ADPCM) that splits the frequency band into two sub-bands (0 to 4000 Hz and 4000 to 8000 Hz) and applies ADPCM to sub-bands independently. G.722 operates in three compression rates. The recommended default bit rate of 64 kbps gives the highest quality. A compression rate of 56 kbps allows for an auxiliary data channel of 8 kbps, and the 48-kbps mode allows for an auxiliary data channel of 16 kbps. Because of the availability of high-throughput independent data ...