Speech processing system by use of short-time crosscorrelation function

SPAC is a new speech processing system for the purposes of expansion or compression of speech spectrum, of shortening or lengthening of speech signal and of improving the signal to noise ratio. A defect of SPAC is generation of squaring distortion which enhances the dynamic range of speech in the time and the spectrum domain. A complicated equalizer is required for compensation of the distortion. This paper proposes a method named SPOC to solve this problem by use of short-time crosscorrelation function of input waveform and its dichotomized wave-form instead of autocorrelation function employed in SPAC. Pair comparison tests are carried out on speech processed by SPAC and SPOC. It is revealed that the quality of SPOC is slightly poorer than that of SPAC, but SPOC is superior to SPAC for the purpose of noise reduction.