Speech perception in noise with binary gains

Abstract

For a given mixture of speech and noise, an ideal binary time-frequency mask is constructed by whether SNR within individual time-frequency units exceeds a local SNR criterion (LC). With linear filters, co-reducing mixture SNR and LC does not alter the ideal binary mask. Taking this manipulation to the limit by setting both mixture SNR and LC to minus infinity produces an output that contains only noise with no target speech at all. This particular output corresponds to turning on or off the filtered noise according to a pattern prescribed by the ideal binary mask. Our study was designed to test on speech intelligibility of noise gated by the ideal binary mask obtained this way. It is observed that listeners achieve nearly perfect speech recognition from gated noise. Only sixteen filter channels and a frame rate of one hundred Hertz 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 in noise.

Speech-cue transmission by an algorithm to increase consonant recognition in noise for hearing-impaired listeners

“…As with the algorithm, it was found that the IBM improved recognition of isolated phonemes, thus extending results observed previously for sentences (Anzalone et al, 2006;Brungart et al, 2006;Li and Loizou, 2008;Wang et al, 2008Wang et al, , 2009Cao et al, 2011;Sinex, 2013). This result should not be surprising, given the effectiveness of algorithm processing observed here and the fact that the algorithm aimed to estimate the IBM.…”

“…In the Ideal Binary Mask (IBM), the premixed speech and noise signals are known, and the mask corresponds to ideal classification. IBM processing produces remarkable speech-intelligibility improvements in noise for both HI and NH listeners, even at extremely low SNRs (Anzalone et al, 2006;Brungart et al, 2006;Li and Loizou, 2008;Wang et al, 2008Wang et al, , 2009Cao et al, 2011;Sinex, 2013).…”

Requirements for the evaluation of computational speech segregation systems

“…The IBM is a binary matrix that classifies the timefrequency (T-F) representation of noisy speech into target-dominated (reliable) and masker-dominated (unreliable) T-F units. Several studies have shown that the IBM can be used to reconstruct a target signal from a noisy mixture, resulting in large improvements of speech intelligibility (e.g., Brungart et al, 2006;Wang et al, 2008). However, the construction of the IBM requires a priori knowledge about the target and the masker, which is typically not available in practice.…”

Evaluating Source Separation Algorithms With Reverberant Speech

“…Quality is a property of speech that corresponds to its realism and naturalness, characteristics that are not necessary for intelligibility. [3] and [7] have shown that a frequency-dependent gating or modulation of noise, which has low quality, can be highly intelligible. Thus, while ASR performance can be predicted by quality to some extent, the relationship is imperfect and indirect [8], [9].…”

“…While human intelligibility has been used to evaluate the quality of ground truth masking-based source separation [2], [3], such evaluations are expensive and time consuming and must be rerun for every variant of a source separation algorithm. ASR performance, on the other hand, requires only computational resources.…”