Speech intelligibility of Chinese time-reversed speech

This study investigated speech intelligibility of Chinese time-reversed speech in a psychoacoustic experiment with different frame lengths of time reversal window. The test of speech intelligibility showed that the intelligibility was high when the frame length was below 64 ms, the intelligibility reduced gradually when the frame length was from 64 to 203 ms, and the intelligibility nearly got to zero when the frame length was above 203 ms. The intelligibility with the frame length 8 ms reduced due to the tonal distortion. The modulation spectra of the original speech and the corresponding time-reversed speech were analyzed and it showed that the intelligibility was correlated with modulation spectra distortion. Therefore, the modulation spectra distortion was conducted by normalizing correlation between the narrow-band envelopes of the original speech and the corresponding time-reversed speech. The objective values were calculated by the speech-based speech transmission index method and it showed that the objective values were highly correlated with the test of speech intelligibility (r=0.876, p < 0.01). The study demonstrates that speech intelligibility is related to narrow-band envelopes and the preservation of narrow-band envelopes is correlated with the intelligibility of time-reversed speech.