An analysis/synthesis system based on the sinusoidal speech model has been developed that produces synthetic speech that is perceptually equivalent to the original: A zero-phase harmonic version of this system, using only pitch, voicing and amplitude information, has achieved a Diagnostic Aceptibility Measure (DAM) of 63.0. This shows that high-quality synthetic speech can be achieved at low rates provided the sine-wave amplitude information is coded efficiently. One approach has been to linearly interpolate the sine-wave amplitudes and then frequency-warp, downsample and code using DPCM quantization. Another has been to fit an all-pole model to the sine-wave amplitudes and code the resulting reflection coefficients. In either case, frequency domain post-filtering is needed to reduce the effects of coder noise. The system, called the Sinusoidal Transform Coder (STC), has been implemented in basic 4800 b/s form using two ADSP 2100 fixed-point DSP chips, with more complex versions at 2400 b/s and 4800 b/s requiring an additional DSP chip. The STC system is designed for, and being evaluated in, application environments requiring robust performance, including military airborne and civilian air traffic control communications systems.
The sinusoidal transform coder (STC): A high-performance multi-rate speech coder
Sprachsynthese mittels STC, einem Sprachcodierer mit hoher Leistung
1989
4 Seiten, 2 Tabellen, 6 Quellen
Aufsatz (Konferenz)
Englisch
Engineering Index Backfile | 1959
A 6.6kb/s CELP Speech Coder: High Performance for GSM Half-Rate System
British Library Conference Proceedings | 1994
|Computationally Efficient Implementation of AMR Speech Coder
British Library Conference Proceedings | 2003
|VLSI Universal Noiseless Coder
NTRS | 1989
|