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

Abstract

Consonant recognition was assessed following extraction of speech from noise using a more efficient version of the speech-segregation algorithm described in Healy, Yoho, Wang, and Wang [(2013) J. Acoust. Soc. Am. 134, 3029-3038]. Substantial increases in recognition were observed following algorithm processing, which were significantly larger for hearing-impaired (HI) than for normalhearing (NH) listeners in both speech-shaped noise and babble backgrounds. As observed previously for sentence recognition, older HI listeners having access to the algorithm performed as well or better than young NH listeners in conditions of identical noise. It was also found that the binary masks estimated by the algorithm transmitted speech features to listeners in a fashion highly similar to that of the ideal binary mask (IBM), suggesting that the algorithm is estimating the IBM with substantial accuracy. Further, the speech features associated with voicing, manner of articulation, and place of articulation were all transmitted with relative uniformity and at relatively high levels, indicating that the algorithm and the IBM transmit speech cues without obvious deficiency. Because the current implementation of the algorithm is much more efficient, it should be more amenable to real-time implementation in devices such as hearing aids and cochlear implants.

Improving competing voices segregation for hearing impaired listeners using a low-latency deep neural network algorithm

“…It is important to find the most relevant way of assessing the benefit from such separation algorithms, and even though many studies use objective metrics, such as the STOI and ESTOI (Jensen and Taal, 2016;Taal et al, 2011), the final assessment should be a speech recognition test on the target group of listeners. Earlier work on separation has used consonant recognition in hearing-impaired listeners as an outcome measure and found a benefit in both speech shaped and babble noises (Healy et al, 2014). This type of benefit was also confirmed in novel noise types, which is a basic requirement for successful application of such an algorithm (Healy et al, 2015).…”

An algorithm to increase speech intelligibility for hearing-impaired listeners in novel segments of the same noise type

“…As described below, the algorithm tested currently differs from those employed in Healy et al (2013) and Healy et al (2014) in several aspects. Whereas the goal of the algorithms employed by Healy et al (2013) and Healy et al (2014) was to estimate the IBM, the current algorithm estimates the Ideal Ratio Mask (IRM; Srinivasan et al, 2006;Narayanan and Wang, 2013;Wang et al, 2014). To address the challenge of unseen noise segments, the current algorithm was trained using substantially longer noises, which were further expanded using a noise-perturbation technique (Chen et al, 2015).…”

“…A subsequent study (Healy et al, 2014) involved recognition of isolated consonants in order to identify the specific speech cues transmitted by the algorithm and the IBM. Consonant recognition in speech-shaped noise and babble was substantially increased by the algorithm for both NH and HI listeners, despite the lack of top-down cues associated with sentence recognition and the correspondingly increased reliance on bottom-up acoustic cues.…”

“…The use of the same brief noise segment has the potential to substantially increase "overfit" to the training conditions, thus limiting generalizability. Healy et al (2013) and Healy et al (2014) used longer-duration noise (10 s) with looping, in an effort to decrease overfit and increase the potential to generalize. However, the background noise selected for each utterance was drawn from the same overall 10-s noise segment during both algorithm training and operation.…”

Speech enhancement based on neural networks improves speech intelligibility in noise for cochlear implant users

“…Using a similar approach, algorithms have been trained on labelled datasets to approximate the IBM. These have been reported to provide remarkably large SI improvements for NH listeners (Kim et al., 2009), HI-listeners (Healy et al., 2013, Healy et al., 2014) and CI users (Hu and Loizou, 2010) for speech in both stationary and non-stationary noise, even at low SNRs. However, these algorithms were trained and tested on datasets using the same speaker, background noise and SNRs.…”