Voice Signal Processing in GSM


The radio channel is quite different from the wired channel. First, the radio channel has a distinct time-change characteristic. The radio channel is exposed to the air, so it is vulnerable to the  interferences in the air. The signal is influenced by various interferences, multi-path fading and shadow fading, so the error bit ratio is rather high. To solve the problems mentioned above, a series of forward and backward(uplink & downlink) transmission techniques are applied. The original subscriber data or signaling data are transformed before being carried by the radio waves. And at the other end of the transmission, a reverse transforming will be done. This can provide necessary protection to the transmitting signal. The transformation methods roughly include the channel coding/decoding, interleaving/de-interleaving, burst formatting, encryption/decryption, and modulation/demodulation. For the voice, to pass an analog-to-digital converter is actually a sampling process in the rate of 8KHz,after quantification each 125μs contains 13bit of code stream; then speech coding is performed  with every 20ms as a segment and the code transmission rate is reduced to 13Kbit/s, which becomes 22.8Kbit/s after the channel coding; then the voice becomes a code stream at 33.8kbit/s after code interleaving, encryption and burst formatting and is transmitted finally. The processing at the terminal is just the reverse of the above procedures.