Packet Loss Concealment - Zero Insertion - Waveform Substitution - Model-based Methods

The animated tutorial can be found here:
packet-loss-concealmentfla.swf

Quick Review - Analog samples are placed into digital quantities (quantizing) at specific intervals which means logarithmic (step) quantizing is not a linear process or not all analog values are sampled. However, there are enough samples to reproduce voice but not hi-fidelity sounds. That is, different quantizing systems such as µ-Law and A-Law sample voice differently resulting in different reproduction qualities.

While voice frequencies are linear or sequential, human voice has unique characteristics. That is, quantization levels according to a logarithmic, instead of a linear, process provides smaller quantization levels at lower signal often lower frequency - bass voice. The result is a higher level of reproduction for human voice. A VOCODER-VOice-CODER device allows the DSP-Digital Signal Processor to amplify (quantize) lower or higher frequencies based on the application. That is, the VOCODER can change the logarithms for processing the speech or other input signal such as music. Such a device is also called a COMPANDER-COMpressor and exPANDer. In other words, different Companders will process voice samples differently. The North American PCM-Pulse Code Modulation is called µ-Law (µ from the Greek mu or M or Modulation) which approximates the logarithmic (steps) curve. Mu-Law encoding is where the steps within a cord are all the same size and therefore linear.

PLC-Packet Loss Concealment are advanced voice processing techniques designed to overcome packet loss, bit loss and packet arrival losses or delays. Packets may arrive out-of-order, have excessive delays or be unrecoverable/lost. There are three of more common tools that reduce voice loss or distortion; they include zero insertion (also called zero substitution), waveform substitution replacement and model-based methods.
1 - Zero substitution is simply the replacement of lost bits with zero bits to fill the byte or partial byte lost.
2 - Waveform substitution shown here replaces the bytes lost with those of the previous byte stream (like an instant replay).
3 - Model-based methods use expert-systems advanced voice techniques such as those found in VOCODERS to replace speech lost (e.g. remove syllables) or removed (e.g. no signal "dead air").
According to Alan Percy Director of Market Development for AudioCodes (www.audiocodes.com), on the issue of Packet Loss Concealment, "AudioCodes newest voice coder EG.711 implementation is more aggressive with packet loss concealment and can handle quite high packet loss while still being understandable."

This presentation is also included in TMC University special course on Microsoft OCS-Office Communications Server at ITexpo.com. For more go here: http://www.tmcnet.com/voip/conference/west-08/tmc-university-microsoft-ocs.htm

This presentation is also included in online/onsite courses SIP 2.0c and for OCS-101 Office Communications Server per person (volume and site license discounts available). Discounts are also available to members of the SIP Forum and MS Partners. For customizing, special discounts, website animations, technical/sales training, technical writing and other services, go to http://www.techtionary.com or please call Tom Cross at 303-594-1694 or cross@gocross.com.