Search

Highlights • Spectral and temporal modification techniques combine synergistically. • For Spanish sentences, error rates are reduced by a factor of 3 compared to unmodified speech. • All phonemes benefit from spectral and temporal modification. • A glimpsing model predicts listener performance with a correlation of 0.96. • Cochlear-Scaled Entropy does not improve the performance of a retiming algorithm. Modifying clean speech prior to output in noisy conditions can lead to substantial intelligibility gains. Most algorithms operate by redistributing energy across the signal, leaving the timing of the underlying speech sounds intact. Other techniques do alter the timing of speech relative to the masker. Both classes of approach – spectral and temporal – lead to a reduction in energetic masking. The current study examines how their combination affects intelligibility. Arguments can be made for both synergy and redundancy, and the presence of distortions introduced by both spectral and temporal approaches might even lead to an antagonistic combination. A cohort of native Spanish listeners identified keywords in sentences in unmodified form and following spectral, temporal and spectro-temporal modification, in the presence of a fluctuating masker. Errors in the spectro-temporal condition were substantially lower than following spectral or temporal modification alone, with a three-fold reduction compared to unmodified speech. Spectro-temporal gains were observed for all phonemes. A glimpse-based model of energetic masking incorporating speech rate changes predicts intelligibility (r =. 96), and a glimpsing analysis provides further insights into the distinct mechanisms through which spectral and temporal approaches lead to a release from energetic masking. [ABSTRACT FROM AUTHOR]

Near-end speech enhancement works by modifying speech prior to presentation in a noisy environment, typically operating under a constraint of limited or no increase in speech level. One issue is the extent to which near-end enhancement techniques require detailed estimates of the masking environment to function effectively. The current study investigated speech modification strategies based on reallocating energy statically across the spectrum using masker-specific spectral weightings. Weighting patterns were learned offline by maximising a glimpse-based objective intelligibility metric. Keyword scores in sentences in the presence of stationary and fluctuating maskers increased, in some cases by very substantial amounts, following the application of masker- and SNR-specific spectral weighting. A second experiment using generic masker-independent spectral weightings that boosted all frequencies above 1 kHz also led to significant gains in most conditions. These findings indicate that energy-neutral spectral weighting is a highly-effective near-end speech enhancement approach that places minimal demands on detailed masker estimation. [ABSTRACT FROM AUTHOR]

Several modification algorithms that alter natural or synthetic speech with the goal of improving intelligibility in noise have been proposed recently. A key requirement of many modification techniques is the ability to predict intelligibility, both offline during algorithm development, and online, in order to determine the optimal modification for the current noise context. While existing objective intelligibility metrics (OIMs) have good predictive power for unmodified natural speech in stationary and fluctuating noise, little is known about their effectiveness for other forms of speech. The current study evaluated how well seven OIMs predict listener responses in three large datasets of modified and synthetic speech which together represent 396 combinations of speech modification, masker type and signal-to-noise ratio. The chief finding is a clear reduction in predictive power for most OIMs when faced with modified and synthetic speech. Modifications introducing durational changes are particularly harmful to intelligibility predictors. OIMs that measure masked audibility tend to over-estimate intelligibility in the presence of fluctuating maskers relative to stationary maskers, while OIMs that estimate the distortion caused by the masker to a clean speech prototype exhibit the reverse pattern. [ABSTRACT FROM AUTHOR]

Highlights: [•] Speech production is affected by environmental and interlocutor-related contextual factors. [•] We review speech modifications produced by talkers in response to context. [•] Algorithmic modifications aimed at increasing intelligibility are also reviewed. [•] A summary table of 46 candidate speech modifications is provided. [ABSTRACT FROM AUTHOR]

Abstract: Robust speech recognition in everyday conditions requires the solution to a number of challenging problems, not least the ability to handle multiple sound sources. The specific case of speech recognition in the presence of a competing talker has been studied for several decades, resulting in a number of quite distinct algorithmic solutions whose focus ranges from modeling both target and competing speech to speech separation using auditory grouping principles. The purpose of the monaural speech separation and recognition challenge was to permit a large-scale comparison of techniques for the competing talker problem. The task was to identify keywords in sentences spoken by a target talker when mixed into a single channel with a background talker speaking similar sentences. Ten independent sets of results were contributed, alongside a baseline recognition system. Performance was evaluated using common training and test data and common metrics. Listeners’ performance in the same task was also measured. This paper describes the challenge problem, compares the performance of the contributed algorithms, and discusses the factors which distinguish the systems. One highlight of the comparison was the finding that several systems achieved near-human performance in some conditions, and one out-performed listeners overall. [Copyright &y& Elsevier]

Abstract: This paper addresses the problem of recognising speech in the presence of a competing speaker. We review a speech fragment decoding technique that treats segregation and recognition as coupled problems. Data-driven techniques are used to segment a spectro-temporal representation into a set of fragments, such that each fragment is dominated by one or other of the speech sources. A speech fragment decoder is used which employs missing data techniques and clean speech models to simultaneously search for the set of fragments and the word sequence that best matches the target speaker model. The paper investigates the performance of the system on a recognition task employing artificially mixed target and masker speech utterances. The fragment decoder produces significantly lower error rates than a conventional recogniser, and mimics the pattern of human performance that is produced by the interplay between energetic and informational masking. However, at around 0dB the performance is generally quite poor. An analysis of the errors shows that a large number of target/masker confusions are being made. The paper presents a novel fragment-based speaker identification approach that allows the target speaker to be reliably identified across a wide range of SNRs. This component is combined with the recognition system to produce significant improvements. When the target and masker utterance have the same gender, the recognition system has a performance at 0dB equal to that of humans; in other conditions the error rate is roughly twice the human error rate. [Copyright &y& Elsevier]