On the limits of speech recognition in noise

Peters, S.D.; Stubley, Peter; Valin, Jean‐Marc

doi:10.1109/icassp.1999.758138

Cited by 19 publications

(14 citation statements)

References 2 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Experiments conducted by Peters et al (1999) demonstrate that these conclusions are not correct in case of noisy speech recordings. He suggests that information lost by the conventional acoustic analysis (phase and fine spectral resolution) may become crucial for intelligibility in case of speech distortions (reverberation, environment noise, etc.).…”

Section: Specific Methodologiesmentioning

confidence: 94%

See 1 more Smart Citation

Automatic speech recognition and speech variability: A review

BenZeghiba

Mori

Deroo

et al. 2007

Speech Communication

431

207

View full text Add to dashboard Cite

Major progress is being recorded regularly on both the technology and exploitation of automatic speech recognition (ASR) and spoken language systems. However, there are still technological barriers to flexible solutions and user satisfaction under some circumstances. This is related to several factors, such as the sensitivity to the environment (background noise), or the weak representation of grammatical and semantic knowledge.Current research is also emphasizing deficiencies in dealing with variation naturally present in speech. For instance, the lack of robustness to foreign accents precludes the use by specific populations. Also, some applications, like directory assistance, particularly stress the core recognition technology due to the very high active vocabulary (application perplexity). There are actually many factors affecting the speech realization: regional, sociolinguistic, or related to the environment or the speaker herself. These create a wide range of variations that may not be modeled correctly (speaker, gender, speaking rate, vocal effort, regional accent, speaking style, non-stationarity, etc.), especially when resources for system training are scarce. This paper outlines current advances related to these topics.

show abstract

Section: Specific Methodologiesmentioning

confidence: 94%

“…It has been suggested (Demuynck et al, 2004;Leonard, 1984;Peters et al, 1999) that conventional cepstral representation of speech may destroy important information by ignoring the phase (power spectrum estimation) and reducing the spectral resolution (Mel filter bank, LPC, cepstral liftering, etc. ).…”

Section: Specific Methodologiesmentioning

confidence: 99%

Automatic speech recognition and speech variability: A review

BenZeghiba

Mori

Deroo

et al. 2007

Speech Communication

431

207

View full text Add to dashboard Cite

show abstract

“…To account for this uncertainty in the probability that communication actually occurs we weight the SE values throughout the area by a probability-of-recognition term, PR; where PR = 0.5 at SE = 0 dB, and PR = 1.0 at SE = 18 dB. The upper value of 18 dB is assumed based on analogy to recognition thresholds in human speech (Pearsons et al 1977, Tafalla & Evans 1997, Peters et al 1999. We refer to this area weighted by a recognition function as a sender communication space.…”

Section: Communication Masking Terms and Algorithmmentioning

confidence: 99%

Acoustic masking in marine ecosystems: intuitions, analysis, and implication

Clark¹,

Ellison²,

Southall³

et al. 2009

Mar. Ecol. Prog. Ser.

488

369

View full text Add to dashboard Cite

Acoustic masking from anthropogenic noise is increasingly being considered as a threat to marine mammals, particularly low-frequency specialists such as baleen whales. Low-frequency ocean noise has increased in recent decades, often in habitats with seasonally resident populations of marine mammals, raising concerns that noise chronically influences life histories of individuals and populations. In contrast to physical harm from intense anthropogenic sources, which can have acute impacts on individuals, masking from chronic noise sources has been difficult to quantify at individual or population levels, and resulting effects have been even more difficult to assess. This paper presents an analytical paradigm to quantify changes in an animal's acoustic communication space as a result of spatial, spectral, and temporal changes in background noise, providing a functional definition of communication masking for free-ranging animals and a metric to quantify the potential for communication masking. We use the sonar equation, a combination of modeling and analytical techniques, and measurements from empirical data to calculate time-varying spatial maps of potential communication space for singing fin (Balaenoptera physalus), singing humpback (Megoptera novaeangliae), and calling right (Eubalaena glacialis) whales. These illustrate how the measured loss of communication space as a result of differing levels of noise is converted into a time-varying measure of communication masking. The proposed paradigm and mechanisms for measuring levels of communication masking can be applied to different species, contexts, acoustic habitats and ocean noise scenes to estimate the potential impacts of masking at the individual and population levels.

show abstract

“…The recognition accuracy based on the majority of three listeners was 99.9%, indicating that signals resynthesized from the spectral envelope of short-time fragments of speech are sufficient in acoustically optimal conditions. Peters et al (1999) carried out a comparison of HSR and ASR recognition performance with unaltered and resynthesized speech. Feature vectors calculated from noisy digits were converted to audible signals based on an analytical processing scheme.…”

Section: Introductionmentioning

confidence: 99%

Effect of speech-intrinsic variations on human and automatic recognition of spoken phonemes

Meyer

Brand

Kollmeier

2011

The Journal of the Acoustical Society of America

View full text Add to dashboard Cite

The aim of this study is to quantify the gap between the recognition performance of human listeners and an automatic speech recognition (ASR) system with special focus on intrinsic variations of speech, such as speaking rate and effort, altered pitch, and the presence of dialect and accent. Second, it is investigated if the most common ASR features contain all information required to recognize speech in noisy environments by using resynthesized ASR features in listening experiments. For the phoneme recognition task, the ASR system achieved the human performance level only when the signal-to-noise ratio (SNR) was increased by 15 dB, which is an estimate for the human-machine gap in terms of the SNR. The major part of this gap is attributed to the feature extraction stage, since human listeners achieve comparable recognition scores when the SNR difference between unaltered and resynthesized utterances is 10 dB. Intrinsic variabilities result in strong increases of error rates, both in human speech recognition (HSR) and ASR (with a relative increase of up to 120%). An analysis of phoneme duration and recognition rates indicates that human listeners are better able to identify temporal cues than the machine at low SNRs, which suggests incorporating information about the temporal dynamics of speech into ASR systems.

show abstract

On the limits of speech recognition in noise

Cited by 19 publications

References 2 publications

Automatic speech recognition and speech variability: A review

Automatic speech recognition and speech variability: A review

Acoustic masking in marine ecosystems: intuitions, analysis, and implication

Effect of speech-intrinsic variations on human and automatic recognition of spoken phonemes

Contact Info

Product

Resources

About