Methods for analysis and synthesis of speech directly in the time domain (for purposes of time or bandwidth compression) are discussed. Specifically, autocorrelation and cross-correlation are considered and -time-domain vocoders- based on these principles of analysis are described. The spectrum squaring inherent in autocorrelation analysis is corrected for by a special time-varying equalizer. The cross-correlation is performed between speech and a speech-derived signal with a flat spectral envelope. Both autocorrelation and cross-correlation vocoders were simulated on a digital computer (IBM 7090) using special programming techniques. The remade speech quality was judged superior or equal to the best known spectrum channel vocoders for comparable bandwidth, bandwidth compression, and synthesizer excitation.
Click to purchase paper as a non-member or login as an AES member. If your company or school subscribes to the E-Library then switch to the institutional version. If you are not an AES member and would like to subscribe to the E-Library then Join the AES!
This paper costs $33 for non-members and is free for AES members and E-Library subscribers.