Speech coding at 9.6 Kb/s and below using vector quantized transform coder

The vector-quantized transform coder (VQTC) is proposed as an alternative to a code-excited linear predicting (CELP) coder. The consideration of more efficient and subjectively meaningful residual quantization allows the VQTC to operate at rates as low as 6 kb/s without significant degradations in speech quality. Below 6 kb/s the quality of the speech was good and did not contain any background noise peculiar to transform coders. Simulations of the VQTC and standard CELP were performed to assess the performance of the VQTC as an alternative coding scheme for low bit rates. In general, the performance of the VQTC was as good as CELP at rates of 9 to 4.8 kb/s. However, the VQTC requires substantially fewer computations than CELP, which makes the VQTC attractive for real-time applications.<
>