Some acoustic-phonetic correlates of speech produced in noise

Pisoni, David B.; Bernacki, Robert H.; Nusbaum, Howard C.; Yuchtman, Moshe

doi:10.1109/icassp.1985.1168217

Cited by 37 publications

(33 citation statements)

References 5 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…After a slight manipulation and substituting from (12), the first inequality (17) can be rewritten (18) where and are non-negative constants representing the relation between the clean speech and noise power spectral means and variances: , . The inequality is true for all and such that (19) The second inequality in (17) results in an identical expression. Clearly, all non-negative satisfy inequality (19), which means that if the ratio between the means of noise and clean speech power spectrum bins, , is bigger than the ratio between their variances, , the variance of the noisy speech log power spectrum bin will always decrease, with a rate inversely proportional to SNR.…”

Section: A Speech Coding In Noisementioning

confidence: 64%

“…In the spectral domain, the energy of LE waveforms migrates to higher frequencies, resulting in an upward shift of spectral center of gravity [4], [7], and in flattening of the spectral slope of short-time speech spectra [3], [19], [20]. The first formant center frequency varies inversely to the vertical position of the tongue and the second formant frequency increases with tongue advancement [21].…”

mentioning

confidence: 99%

“…In [19], the increases are accompanied by consistent decreases, while in [3] and [25] the locations of both shift up in frequency for most phones. Average bandwidths of the first four formants reduce in LE for most phones [3], [4], [6], [25].…”

mentioning

confidence: 99%

See 2 more Smart Citations

Unsupervised Equalization of Lombard Effect for Speech Recognition in Noisy Adverse Environments

Bořil

Hansen

2010

IEEE Trans. Audio Speech Lang. Process.

View full text Add to dashboard Cite

Abstract-In the presence of environmental noise, speakers tend to adjust their speech production in an effort to preserve intelligible communication. The noise-induced speech adjustments, called Lombard effect (LE), are known to severely impact the accuracy of automatic speech recognition (ASR) systems. The reduced performance results from the mismatch between the ASR acoustic models trained typically on noise-clean neutral (modal) speech and the actual parameters of noisy LE speech. In this study, novel unsupervised frequency domain and cepstral domain equalizations that increase ASR resistance to LE are proposed and incorporated in a recognition scheme employing a codebook of noisy acoustic models. In the frequency domain, short-time speech spectra are transformed towards neutral ASR acoustic models in a maximum-likelihood fashion. Simultaneously, dynamics of cepstral samples are determined from the quantile estimates and normalized to a constant range. A codebook decoding strategy is applied to determine the noisy models best matching the actual mixture of speech and noisy background. The proposed algorithms are evaluated side by side with conventional compensation schemes on connected Czech digits presented in various levels of background car noise. The resulting system provides an absolute word error rate (WER) reduction on 10-dB signal-to-noise ratio data of 8.7% and 37.7% for female neutral and LE speech, respectively, and of 8.7% and 32.8% for male neutral and LE speech, respectively, when compared to the baseline recognizer employing perceptual linear prediction (PLP) coefficients and cepstral mean and variance normalization.Index Terms-Cepstral compensation, codebook of noisy models, frequency warping, Lombard effect, speech recognition.

show abstract

Section: A Speech Coding In Noisementioning

confidence: 64%

mentioning

confidence: 99%

See 1 more Smart Citation

Unsupervised Equalization of Lombard Effect for Speech Recognition in Noisy Adverse Environments

Bořil

Hansen

2010

IEEE Trans. Audio Speech Lang. Process.

View full text Add to dashboard Cite

show abstract

“…In the perceptual domain, intelligibility studies established that naturally produced clear speech enhances intelligibility for various listener populations including adults with normal or impaired hearing [7,8,17,19,26,27] elderly adults [23,9] both native and non-native listeners of the target language [2,4,25] and children with and without learning impairments [3].…”

Section: Introductionmentioning

confidence: 99%

Intelligibility and Acoustic Characteristics of Clear and Conversational Speech in Telugu (A South Indian Dravidian Language)

Durisala¹,

Prakash

Nambi

et al. 2011

Indian J Otolaryngol Head Neck Surg

View full text Add to dashboard Cite

The overall goal of this study is to examine the intelligibility differences of clear and conversational speech and also to objectively analyze the acoustic properties contributing to these differences. Seventeen postlingual stable sensory-neural hearing impaired listeners with an age range of 17-40 years were recruited for the study. Forty Telugu sentences spoken by a female Telugu speaker in both clear and conversational speech styles were used as stimuli for the subjects. Results revealed that mean scores of clear speech were higher (mean = 84.5) when compared to conversational speech (mean = 61.4) with an advantage of 23.1% points. Acoustic properties revealed greater fundamental frequency (f0) and intensity, longer duration, higher consonant-vowel ratio (CVR) and greater temporal energy in clear speech.

show abstract

“…Other factors are para-linguistic; they depend on the speaker's intention, physical condition, speaking style, dialect, emphasis, and emotion. The Lombard effect in noisy environments [69] belongs to this category. Although the intrinsic causes are more important, it is difficult to separate them from the other causes.…”

Section: Introductionmentioning

confidence: 99%

Acoustic Model Adaptation for Speech Recognition

Sairyo

2010

IEICE Trans. Inf. & Syst.

View full text Add to dashboard Cite

SUMMARYStatistical speech recognition using continuous-density hidden Markov models (CDHMMs) has yielded many practical applications. However, in general, mismatches between the training data and input data significantly degrade recognition accuracy. Various acoustic model adaptation techniques using a few input utterances have been employed to overcome this problem. In this article, we survey these adaptation techniques, including maximum a posteriori (MAP) estimation, maximum likelihood linear regression (MLLR), and eigenvoice. We also present a schematic view called the adaptation pyramid to illustrate how these methods relate to each other. key words: speech recognition, acoustic model adaptation, hidden Markov models

show abstract

Some acoustic-phonetic correlates of speech produced in noise

Cited by 37 publications

References 5 publications

Unsupervised Equalization of Lombard Effect for Speech Recognition in Noisy Adverse Environments

Unsupervised Equalization of Lombard Effect for Speech Recognition in Noisy Adverse Environments

Intelligibility and Acoustic Characteristics of Clear and Conversational Speech in Telugu (A South Indian Dravidian Language)

Acoustic Model Adaptation for Speech Recognition

Contact Info

Product

Resources

About