A real time processing TCELP coder/decoder at 4.8 kbit/sec using LSP split vector quantization

Before transmission in a narrow band channel, the speech signal has to be compressed. The Code Excited Linear Prediction Coder (CELP) makes it possible to synthesize good quality speech at low bit rates. Different speech linear predictive coding parameters can be used to design the speech spectral envelope. In our coder the Line Spectrum Pairs (LSP), belonging to the frequency domain, enable us to design the vocal track transfer function. In its first version the coder uses a CLSP (LSP cosine function) scalar quantization leading to a rate of 5.45 kbit/sec. In order to reduce the bit rate to a standard (4.8 kbit/sec), vector quantization was introduced in our coder.