Applied mel-frequency discrete wavelet coefficients and parallel model compensation for noise-robust speech recognition

Tüfekçi, Zekeriya; Gowdy, J.N.; Gurbuz, Sabri; Patterson, Eric

doi:10.1016/j.specom.2006.06.006

Cited by 29 publications

(12 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Instead of using the MFCCs, other types of features are proposed by the researchers to increase the robustness of the recognizers [4][5][6][7][8]. Also, since the MFCCs are widely adopted, many researchers have made effort to improve its robustness under noise by modifying, or changing, some processes in the conventional scheme [9][10][11][12][13].…”

Section: Introductionmentioning

confidence: 99%

A robust polynomial regression-based voice activity detector for speaker verification

DİŞKEN

Tüfekçi

Çevik

2017

J AUDIO SPEECH MUSIC PROC.

View full text Add to dashboard Cite

Robustness against background noise is a major research area for speech-related applications such as speech recognition and speaker recognition. One of the many solutions for this problem is to detect speech-dominant regions by using a voice activity detector (VAD). In this paper, a second-order polynomial regression-based algorithm is proposed with a similar function as a VAD for text-independent speaker verification systems. The proposed method aims to separate steady noise/silence regions, steady speech regions, and speech onset/offset regions. The regression is applied independently to each filter band of a mel spectrum, which makes the algorithm fit seamlessly to the conventional extraction process of the mel-frequency cepstral coefficients (MFCCs). The kmeans algorithm is also applied to estimate average noise energy in each band for spectral subtraction. A pseudo SNR-dependent linear thresholding for the final VAD output decision is introduced based on the k-means energy centers. This thresholding considers the speech presence in each band. Conventional VADs usually neglect the deteriorative effects of the additive noise in the speech regions. Contrary to this, the proposed method decides not only for the speech presence, but also if the frame is dominated by the speech, or the noise. Performance of the proposed algorithm is compared with a continuous noise tracking method, and another VAD method in speaker verification experiments, where five different noise types at five different SNR levels were considered. The proposed algorithm showed superior verification performance both with the conventional GMM-UBM method, and the stateof-the-art i-vector method.

show abstract

Section: Introductionmentioning

confidence: 99%

A robust polynomial regression-based voice activity detector for speaker verification

DİŞKEN

Tüfekçi

Çevik

2017

J AUDIO SPEECH MUSIC PROC.

View full text Add to dashboard Cite

show abstract

“…Also, Wavelet transform is used for speech enhancement [14]. The implementation of wavelet transform is done using successive digital filters [13] which can be imitated to implement a Mel scale like filter bank for speech recognition [15]. The conventional filter banks of speech recognition such as Mel scale and Bark scale are implemented using wavelet transform instead of Fourier transform [16][17][18].…”

Section: Introductionmentioning

confidence: 99%

Modified Multivariate Euclidean Dynamic Time Warping Based Spoken Keyword Detection

Alex¹,

Nithya²

2017

IJIES

View full text Add to dashboard Cite

Traditional Dynamic Time Warping (DTW) technique find similarities between two one-dimensional time series sequence. Initially, in earlier decades, DTW was not preferred because of its computational complexity. However, due to the evolution of computing power, this has been revisited for spoken keyword detection recently. Conventional spectral features such as Mel-Frequency Cepstral Coefficients (MFCC) and contemporary wavelet features are multi-dimensional in nature which are used in speech recognition. In this work, a new strategy of DTW is proposed to work with multi-dimensional feature vector in calculating the local distance matrix. Additionally, a faster approach is specified to find the similarity in the global distance matrix. The proposed methods are evaluated with MFCC and wavelet features on a connected TIDIGITS corpus for spoken keyword detection system. Experimental results prove that there is an improvement in reduction of computational complexity compared to traditional DTW. Also, contemporary wavelet feature based spoken keyword detection system gave better detection accuracy than MFCC based spoken keyword detection system in the noisy environment.

show abstract

“…Many different approaches have been studied to decrease the effect of noise on the recognition [1]. Wavelet denoising can be applied as a preprocessing stage before feature extraction to compensate noise effects [2].…”

mentioning

confidence: 99%

“…Eventually, many works employ wavelet features in an ASR while such features are computed based on critically sampled filter bands using WT or WP analysis [1,[5][6][7]. The usage of wavelet-based features extracted from the WPD leads to improvement of recognition rate compared with the well-known conventional MFCC features [8] and [9].…”

mentioning

confidence: 99%

Speech Recognition

Farouk

2013

SpringerBriefs in Electrical and Computer Engineering

View full text Add to dashboard Cite

show abstract

Applied mel-frequency discrete wavelet coefficients and parallel model compensation for noise-robust speech recognition

Cited by 29 publications

References 23 publications

A robust polynomial regression-based voice activity detector for speaker verification

A robust polynomial regression-based voice activity detector for speaker verification

Modified Multivariate Euclidean Dynamic Time Warping Based Spoken Keyword Detection

Speech Recognition

Contact Info

Product

Resources

About