In the development process of noise-reduction algorithms, an objective machine-driven intelligibility measure which shows high correlation with speech intelligibility is of great interest. Besides reducing time and costs compared to real listening experiments, an objective intelligibility measure could also help provide answers on how to improve the intelligibility of noisy unprocessed speech. In this paper, a short-time objective intelligibility measure (STOI) is presented, which shows high correlation with the intelligibility of noisy and time-frequency weighted noisy speech (e.g., resulting from noise reduction) of three different listening experiments. In general, STOI showed better correlation with speech intelligibility compared to five other reference objective intelligibility models. In contrast to other conventional intelligibility models which tend to rely on global statistics across entire sentences, STOI is based on shorter time segments (386 ms). Experiments indeed show that it is beneficial to take segment lengths of this order into account. In addition, a free Matlab implementation is provided.

An Algorithm for Intelligibility Prediction of Time–Frequency Weighted Noisy Speech

Existing objective speech-intelligibility measures are suitable for several types of degradation, however, it turns out that they are less appropriate for methods where noisy speech is processed by a time-frequency (TF) weighting, e.g., noise reduction and speech separation. In this paper, we present an objective intelligibility measure, which shows high correlation (rho=0.95) with the intelligibility of both noisy, and TF-weighted noisy speech. The proposed method shows significantly better performance than three other, more sophisticated, objective measures. Furthermore, it is based on an intermediate intelligibility measure for short-time (approximately 400 ms) TF-regions, and uses a simple DFT-based TF-decomposition. In addition, a free Matlab implementation is provided.

/pdf/a-short-time-objective-intelligibility-measure-for-time-5ukywkvokl.pdf

A short-time objective intelligibility measure for time-frequency weighted noisy speech

This paper summarizes twenty studies, published since 1989, that have measured experimentally the relationship between speech recognition in noise and some aspect of cognition, using statistical techniques such as correlation or factor analysis. The results demonstrate that there is a link, but it is secondary to the predictive effects of hearing loss, and it is somewhat mixed across study. No one cognitive test always gave a significant result, but measures of working memory (especially reading span) were mostly effective, whereas measures of general ability, such as IQ, were mostly ineffective. Some of the studies included aided listening, and two reported the benefits from aided listening: again mixed results were found, and in some circumstances cognition was a useful predictor of hearing-aid benefit.

Are individual differences in speech reception related to individual differences in cognitive ability? A survey of twenty experimental studies with normal and hearing-impaired adults

Intelligibility listening tests are necessary during development and evaluation of speech processing algorithms, despite the fact that they are expensive and time consuming. In this paper, we propose a monaural intelligibility prediction algorithm, which has the potential of replacing some of these listening tests. The proposed algorithm shows similarities to the short-time objective intelligibility STOI algorithm, but works for a larger range of input signals. In contrast to STOI, extended STOI ESTOI does not assume mutual independence between frequency bands. ESTOI also incorporates spectral correlation by comparing complete 400ms length spectrograms of the noisy/processed speech and the clean speech signals. As a consequence, ESTOI is also able to accurately predict the intelligibility of speech contaminated by temporally highly modulated noise sources in addition to noisy signals processed with time-frequency weighting. We show that ESTOI can be interpreted in terms of an orthogonal decomposition of short-time spectrograms into intelligibility subspaces, i.e., a ranking of spectrogram features according to their importance to intelligibility. A free MATLAB implementation of the algorithm is available for noncommercial use at http://kom.aau.dk/~jje/.

An Algorithm for Predicting the Intelligibility of Speech Masked by Modulated Noise Maskers

This study which included 23 experienced hearing aid users replicated several of the experiments reported in Gatehouse et al (2003, 2006) with new speech test material, language, and test procedure. The performance measure used was SNR required for 80% correct words in a sentence test. Consistent with Gatehouse et al, this study indicated that subjects showing a low score in a cognitive test (visual letter monitoring) performed better in the speech recognition test with slow time constants than with fast time constants, and performed better in unmodulated noise than in modulated noise, while subjects with high scores on the cognitive test showed the opposite pattern. Furthermore, cognitive test scores were significantly correlated with the differential advantage of fast-acting versus slow-acting compression in conditions of modulated noise. The pure tone average threshold explained 30% of the variance in aided speech recognition in noise under relatively simple listening conditions, while cognitive test scores explained about 40% of the variance under more complex, fluctuating listening conditions, where the pure tone average explained less than 5% of the variance. This suggests that speech recognition under steady-state noise conditions may underestimate the role of cognition in real-life listening.

Interactions between Cognition, Compression, and Listening Conditions: Effects on Speech-in-Noise Performance in a Two-Channel Hearing Aid

The Danish sentence test DANTALE II was developed in analogy to the Swedish sentence test by Hagerman and the German Oldenburg sentence test as a new Danish sentence test in noise to determine the speech reception threshold in noise (SRT, i.e. the signal-to-noise ratio (SNR) that yields 50% intelligibility). Each sentence is generated by a random combination of the alternatives of a base list. This base list consists of 10 sentences with the same syntactical structure (name, verb, numeral, adjective, object). The test sentences were recorded and segmented in such a way that the coarticulation effects were taken into account in order to achieve a high perceived sound quality of the resynthesized sentences: 100 sentences were recorded, each coarticulation between each word and the 10 possible following word alternatives were recorded, and the correct coarticulation was used to generate the test sentences. Word-specific speech recognition curves were measured for each recorded word to optimize the homogeneit...

Jane Lignel Josvassen

Papers

Design, optimization and evaluation of a Danish sentence test in noise.