Speech perception of noise with binary gains

Abstract

For a given mixture of speech and noise, an ideal binary time-frequency mask is constructed by comparing speech energy and noise energy within local time-frequency units. It is observed that listeners achieve nearly perfect speech recognition from gated noise with binary gains prescribed by the ideal binary mask. Only 16 filter channels and a frame rate of 100 Hz are sufficient for high intelligibility. The results show that, despite a dramatic reduction of speech information, a pattern of binary gains provides an adequate basis for speech perception.

Role of mask pattern in intelligibility of ideal binary-masked noisy speech

“…The −60 dB SNR curve, however, is different. First of all, since the mask here was applied to essentially pure noise, this is consistent with the results of Wang et al ͑2008͒ who demonstrated that listeners achieve nearly perfect recognition from IBM-gated noise where the mask is obtained from speech and SSN. This process of producing intelligible speech from noise may be viewed as a form of noise gating.…”

“…In particular, the observation of Wang et al ͑2008͒ that IBM-processed noise is intelligible suggests that the resulting temporal envelope of the processed mixture is important. The speech transmission index ͑Houtgast and Steeneken, 1971͒ considers how distortions to the envelope affect speech intelligibility.…”

“…For example, by taking the mixture SNR to a large negative value, we can measure intelligibility of IBM-gated noise similar to Wang et al ͑2008͒. On the other hand, by taking the mixture SNR to a level near the SRT we may measure how processing with the exact same binary mask affects intelligibility near the SRT.…”

A novel binary mask estimator based on sparse approximation

“…However, one algorithm in particular has shown significant improvements in intelligibility for normal-and impaired-hearing listeners-the ideal binary mask (IBM) [1,2]. The IBM exploits oracle knowledge of the target and interferer signals to preserve only the time-frequency (T-F) regions that are target-dominated.…”

Cepstral smoothing of masks for single-channel speech segregation

“…Similar to [14],which demonstrates that only a few bands of noise modulated by corresponding speech envelope is adequate for speech intelligibility. [11] shows that IBM-gated noise also provides speech intelligibility. In realistic conditions, real noise is never totally stationary, this causes the estimate of IBM error-prone.…”