Frequency-domain spectral envelope estimation for low rate coding of speech

Jelinek, Milan; Adoul, J.-P.

doi:10.1109/icassp.1999.758110

Cited by 11 publications

(4 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The second drawback concerns information extraction to characterize the vocal tract function. It is widely agreed in the speech coding community that it is the spectral envelope and not the gross spectrum that represents the shape of the vocal tract [4]. Although the smoothed spectrum is often similar to the spectral envelope of unvoiced sounds, the situation is quite different in the case of voiced and transitional sounds.…”

Section: A Spectral Envelope Vs Smoothed Spectrummentioning

confidence: 99%

See 1 more Smart Citation

Perceptual harmonic cepstral coefficients as the front-end for speech recognition

Gu¹,

Rose²

2009

J. Acoust. Soc. Am.

View full text Add to dashboard Cite

Perceptual harmonic cepstral coefficients (PHCC) are proposed as features to extract for speech recognition. Pitch estimation and classification into voiced, unvoiced, and transitional speech are performed by a spectro-temporal auto-correlation technique. A peak picking algorithm is then employed to precisely locate pitch harmonics. A weighting function, which depends on the classification and the pitch harmonics, is applied to the power spectrum and ensures accurate representation of the voiced speech spectral envelope. The harmonics weighted power spectrum undergoes mel-scaled band-pass filtering, and the logenergy of the filters' output is discrete cosine transformed to produce cepstral coefficients. For perceptual considerations, within-filter cubic-root amplitude compression is applied to reduce amplitude variation without compromise of the gain invariance properties. Experiments show substantial recognition gains of PHCC over MFCC, with 48% and 15% error rate reduction for the Mandarin digit database and E-set, respectively.

show abstract

Section: A Spectral Envelope Vs Smoothed Spectrummentioning

confidence: 99%

“…In this paper, a new approach is proposed to overcome the above shortcomings, which is inspired by ideas borrowed from speech coding [4]. Rather than average the energy within each filter, the harmonic cepstral coefficients (HCC) are derived for voiced speech from the spectrum envelope sampled at harmonic locations for voiced speech.…”

Section: Introductionmentioning

confidence: 99%

Perceptual harmonic cepstral coefficients as the front-end for speech recognition

Gu¹,

Rose²

2009

J. Acoust. Soc. Am.

View full text Add to dashboard Cite

show abstract

“…4 This assumption is not valid for voiced speech with quasi-periodic excitation. 5 Further, the mean-squared error minimization is for the spectrum of speech, not its envelope, and the spectrum is highly sensitive to noise as well. 5 Wavelet-based features have been proposed for various pattern recognition applications in the field of biomedical signals, 6 microarray data classification, 7 face recognition, 8 and speech recognition application.…”

Section: Introductionmentioning

confidence: 99%

“…5 Further, the mean-squared error minimization is for the spectrum of speech, not its envelope, and the spectrum is highly sensitive to noise as well. 5 Wavelet-based features have been proposed for various pattern recognition applications in the field of biomedical signals, 6 microarray data classification, 7 face recognition, 8 and speech recognition application. [9][10][11][12][13] Wavelet-based features when compared to MFCC features have shown better recognition performance for phoneme recognition 11 isolated digit recognition task 12 and monophone recognition under stressed speech.…”

Section: Introductionmentioning

confidence: 99%

Wavelet Sub-Band Based Temporal Features for Robust Hindi Phoneme Recognition

Farooq

Datta

Shrotriya

2010

Int. J. Wavelets Multiresolut Inf. Process.

View full text Add to dashboard Cite

This paper proposes the use of wavelet transform-based feature extraction technique for Hindi speech recognition application. The new proposed features take into account temporal as well as frequency band energy variations for the task of Hindi phoneme recognition. The recognition performance achieved by the proposed features is compared with the standard MFCC and 24-band admissible wavelet packet-based features using a linear discriminant function based classifier. To evaluate robustness of these features, the NOISEX database is used to add different types of noise into phonemes to achieve signal-to-noise ratios in the range of 20 dB to −5 dB. The recognition results show that under noisy background the proposed technique always achieves a better performance over MFCC-based features.

show abstract