Speech Coder TMS320C5000 DSP Code
ITU G.728 speech coder is intended for telephone-bandwidth signals. The
sample rate on nominally 13-bit samples is 8kHz with the signal bandlimited
to the range 300Hz to 3.4kHz. Compression is by Low Delay Code Excited
Linear Prediction (LD-CELP) to a data rate of 16kb/s.
The principal feature
of G.728 is low throughput delay. Each set of 5 input samples is encoded
to a 10-bit codeword within gain and shape tables. The linear predictive
filter is 50th order and is adapted backwards, enabling the algorithmic
delay to be only 5 samples, ie 0.625ms. Thus in circuit multiplication
over 2-wire analog lines, no echo cancellation is required. Decoded speech
quality is enhanced by the use of a post-filter. G.728 passes low bit
rate modem signals up to 2.4kbit/s, DTMF dialling digits, and network
signalling. It is also a standard for video conferencing audio, and is
often used for speech storage. Mean Opinion Score (MOS) is 3.61.
The Algotron implementation
has been developed to minimise DSP resource usage whilst still passing
the G.728 test vectors. Encoder and decoder are operated as separate items.
With the assembler interface, each has individual 5-sample/codeword subroutine
calls. Further subroutine calls can be made for initialisation. The code
routine calls are re-entrant to allow multi-channel operation, the memory
block associated with each channel being pointer-addressed. A C-callable
wrapper file can also be provided, allowing individual channels to be
called from C for processing of multiple 5-sample/code blocks. A comprehensive
user's guide is provided with the code.
#figures for assembler interfaces with TMS320C54x operation
- program memory figures include read-only tables
NOW - sale is under licence - integration support offered