The decoder portion of the StreamIt GSM Vocoder takes GSM encoded parameters as inputs, and uses these to synthesize audible speech. This is accomplished by processing the parameters through four main filters. The RPE decoder filter produces some "pink noise" that very loosely estimates the speech waveform, using quantized bit sequences and a maximum value parameter from the encoded input. This "pink noise" is fed to the Long Term Prediction portion, which applies long-term characteristics to the sequence through a delay filter within a feedback loop. The resulting signal is then sent to the Short Term Synthesis filter, which decodes high frequency voice characteristics from the encoded parameters and applies these to the signal. Finally, the Post-processing filter identifies peaks in the signal to make it audible. The c reference code for GSM was provided by the Communications and Operating Systems Research Group at the Technische Universität Berlin. (README, COPYRIGHT)