Subjective evaluation of SPAC in improving the quality of noisy speech

Objective and subjective evaluations of SPAC(Speech Processing system by use of Auto-Correlation function) in improving the quality of speech corrupted by additive white noise have shown a fairly good improvement in overall speech quality. The primary objective of this work is to verify quantitatively the subjective speech-to-noise-ratio (SNR) improvement achieved by SPAC. A pair comparison test was conducted to evaluate the SNR improvements for speech signal degraded by additive white noise and coding noise of a low bit-rate ADM codec by comparing to additive and multiplicative noise reference signals respectively. Experimental result shows that SPAC provides the statistically significant SNR improvement for both additive noise and coding noise of ADM at the low quality range. The improvement reaches about 6 dB for -5 dB SNR input having additive noise and about 13 dB for 6.7 kHz ADM which is equivalent to the speech signal of about 0 dB SNR having multiplicative noise.