Microscopic prediction of speech recognition for listeners with normal hearing in noise using an auditory model

Abstract

This study compares the phoneme recognition performance in speech-shaped noise of a microscopic model for speech recognition with the performance of normal-hearing listeners. "Microscopic" is defined in terms of this model twofold. First, the speech recognition rate is predicted on a phoneme-by-phoneme basis. Second, microscopic modeling means that the signal waveforms to be recognized are processed by mimicking elementary parts of human's auditory processing. The model is based on an approach by Holube and Kollmeier [J. Acoust. Soc. Am. 100, 1703-1716 (1996)] and consists of a psychoacoustically and physiologically motivated preprocessing and a simple dynamic-time-warp speech recognizer. The model is evaluated while presenting nonsense speech in a closed-set paradigm. Averaged phoneme recognition rates, specific phoneme recognition rates, and phoneme confusions are analyzed. The influence of different perceptual distance measures and of the model's a-priori knowledge is investigated. The results show that human performance can be predicted by this model using an optimal detector, i.e., identical speech waveforms for both training of the recognizer and testing. The best model performance is yielded by distance measures which focus mainly on small perceptual distances and neglect outliers.

Factors affecting predicted speech intelligibility with cochlear implants in an auditory model for electrical stimulation

“…The proposed model was motivated by a similar approach taken by Jürgens and Brand (2009) and Jürgens et al (2010). Their speech intelligibility prediction was performed using two model stages.…”

Speech recognition error patterns for steady-state noise and interrupted speech

“…While this scoring procedure provides important information about whether a listener recognizes a stimulus item, examining the recognition errors produced by the listener can provide valuable information about how partial information is perceived and resolved by the listener depending on the listening condition. Indeed, there is recent interest in the development of "microscopic" models of speech recognition that take into account fine-grained recognition on a phoneme level (e.g., Cooke, 2006;J€ urgens and Brand, 2009). Toward this end, various corpora have been developed to document consistent word confusions in various types of noise (T oth et al, 2015;Marxer et al, 2016).…”

Refining a model of hearing impairment using speech psychophysics

“…Numerous studies have been reported aiming at predicting the performance of HI listeners in a speech perception task (e.g., J€ urgens and Brand, 2009;Brown et al, 2010), with a peripheral model used as a front end to a conventional automatic speech recognition system, as a back end. A limiting property of such an approach is the inability to decompose the origin of errors, especially front-end versus back-end errors.…”