Performance of transform and subband coding systems combined with harmonic scaling of speech
D. Malah, Ronald E. Crochiere, Richard Vandervoort Cox · The Journal of the Acoustical Society of America · 1980
In this presentation we report on a nearly 2:1 reduction in the bit rate of two high quality waveform encoding schemes—adaptive transform coding (ATC) and subband coding, without impairing the quality of the reconstructed speech. The bit rate reduction is achieved by operating the waveform coders on a frequency scaled speech signal, occupying half of its original band. The frequency scaling is performed by a recently developed efficient time-domain harmonic scaling (TDHS) algorithm. This algorithm is pitch adaptive and is based on weighting adjacent pitch periods by an appropriately chosen window function. Once the pitch period is known, the TDHS algorithm can be implemented with only one multiplication and two additions per output sample. The aim of this study was to achieve high-quality, natural sounding speech at or below 9.6 kbit/s. By combining TDHS with ATC, high-quality encoded speech is demonstrated at 4.8 kbit/s. For high-quality speech at 9.6 kbit/s the combination of TDHS and subband coding offers a system with a lower complexity than ATC and which is potentially amenable for hardware implementation with current technology. Examples of computer processed speech by the above schemes will be played.