VISTA Speech Enhancement System for AM Radio Broadcasting

摘要

A new approach to speech enhancement has been developed for increasing averagetransmission power in the Voice of America broadcast system. The approach uses a sinusoidal analysis/Synthesis framework and integrates phase dispersion, amplitude compression, and spectral shaping to decrease the peak/RMS ratio of the speech waveform so that average transmission power can be increased subject to the peak power limit of the existing transmitters. The processing algorithms adapt dynamically to speech pitch and spectrum, and include phase dispersion technique adapted from radar signal design which minimizes short-time peakiness of the speech waveform while maintaining the original spectral envelope to minimize perceived distortion. Overall, an advantage of about 3 dB in peak/RMS has been achieved relative to commercial devices with quality which has been judged to be comparable for the expected conditions of operational broadcast environments. In order to allow tests of performance in the real broadcast environment and to investigate tradeoffs in degree of peak/RMS reduction versus received speech quality, a real-time prototype has been implemented in the form of a multi-processor based on high-performance digital signal processing chips. The prototype also provides experimental flexibility through control of the degree of processing (allowing mild, normal, and severe), and allows on-line monitoring of the peak/RMS ratio. (RRH)