VoIP Voice and Fax Signal Processing

4.5 POWER-BASED VAD/CNG

Energy- or power-based VAD makes use of 1 byte of payload. The other 10 bytes of G.711 Appendix-II are not available in this energy-based VAD. To maintain the compatibility of the payload, 11 bytes are also used with zero filling of the last 10 bytes. Input to the VAD module is a speech signal. The output from the VAD signifies either the presence or the absence of the speech and 1 byte of power payload at a suitable frame. At the receiver, CNG is generated based on power payload in the VAD packet. The same functional block diagram of Fig. 4.1(e) of VAD/CNG functionality can be referred for power-based VAD/CNG. The power on a frame of samples can be calculated in several ways. Some options used in this book are given here.

The formula given in G.168 [ITU-T-G.168 (2004)] is one of the options to estimate voice power over a block or frame of samples. It is given in Eq. (1.1) in Chapter 1. In VAD-II, the first autocorrelation coefficient R(0) represents power. The parameter R(0) can be calculated easily as sum of squares of amplitudes “e_i” with normalization , where “i” is the index, “k” is the starting sample index of frame, and “n” is the frame length in samples. The power payload has to be suitably normalized to match the calibration of 6.17 dBm matching to 0 dBov.

In practice, VAD detection is set when the VAD power level falls below the −42 ± 2-dBm threshold, ...