Journal Article•
Robust Speech Recognition Based on Vector Taylor Series
TL;DR: In this article, a feature compensation algorithm in the cepstral domain based on vector Taylor series (VTS)expansion is proposed, which can signifi-cantly improve the performance of speech recognition system and outperforms the VTS-based feature compensation method in the log-spectral domain.
Abstract: The vector Taylor series(VTS)expansion is an effective approach to noise robust speech recognition.How-ever,in the log-spectral domain,there exist the strong correlations among the different channels of Mel filter bank and thus it is difficult to estimate the noise variance from noisy speech proposes.A feature compensation algorithm in the cepstral domain based on vector Taylor series was proposed.In this algorithm,the distribution of speech cepstral features was represented by a Gaussian mixture model(GMM),and the mean and variance of noise were estimated from noisy speech by the VTS approximation.The experimental results show that the proposed algorithm can signifi-cantly improve the performance of speech recognition system,and outperforms the VTS-based feature compensation method in the log-spectral domain.
Citations
More filters
TL;DR: This paper studies a speech recognition detection system based on almost interactive reconstruction model, and analyzes the continuous speech signal to show that the proposed model can achieve the better performance than the state-of-the-art methodologies.
Abstract: The natural, fast, stable and reliable interaction between human and machine is the ideal interaction mode pursued by human beings. Speech recognition is a process of pattern matching recognition. Effective speech detection technology can not only reduce the processing time of the system, improve the real-time and accuracy of the system processing, but also eliminate the noise interference of the silent segment, so as to improve the subsequent recognition performance. Through speech recognition, the machine can understand human language, complete the corresponding calculation tasks according to these instructions, and meet the needs of people. The traditional speech recognition method uses acoustic features to describe the model. No matter using model compensation method or normalization method, it can not solve the influence of speaker difference on the performance of recognition system. In this paper, we study a speech recognition detection system based on almost interactive reconstruction model, and analyze the continuous speech signal. The experimental results show that the proposed method has a high recognition rate and a shorter time. Compared with the state-of-the-art methodologies, the proposed model can achieve the better performance.
6 citations
13 Dec 2014
TL;DR: A model adaptation algorithm based on central sub band regression for robust speech recognition, which uses a linear transformation to approximate the relationship between the training and testing conditions for each channel of the Mel filter bank and its adjacent channels is proposed.
Abstract: This paper proposes a model adaptation algorithm based on central sub band regression for robust speech recognition, which uses a linear transformation to approximate the relationship between the training and testing conditions for each channel of the Mel filter bank and its adjacent channels. The maximum likelihood estimation of each channel transform is obtained by several different divisions of all the Mel channels and sub-band adaptation. The experimental results show that the proposed algorithm can obtain more accurate testing acoustic models for rapid model adaptation and outperforms the traditional sub-band regression method.
3 citations
Cites methods from "Robust Speech Recognition Based on ..."
...Generally speaking, robust algorithms for speech recognition may be classified into two major groups: the feature space methods [2-5] and the model adaptation methods [6-10]....
[...]
TL;DR: The result shows that the improved perceptually non-uniform spectral compression feature extraction algorithm can effectively enhance the robustness of speech recognition, and ensure the recognition rate in the noise environments.
2 citations
04 Nov 2016
TL;DR: The research significance of the speech enhancement, the relevant theories of speech signal processing are introduced, and the basic spectral subtraction speech enhancement is expounded, through a lot of simulation experiments verify the effect of spectral subtracted.
Abstract: Abstract In this paper, we present an application of spectral subtraction (SS) algorithm in speech enhancement system to extract the pure speech signal as far as possible. In contrast to the existing research, the proposed algorithm improves the voice quality, which reduces speech distortion, eliminates background noise and improves the speech intelligibility. This paper first introduces the research significance of the speech enhancement, then introduces the relevant theories of speech signal processing, and expounds the basic spectral subtraction speech enhancement, through a lot of simulation experiments verify the effect of spectral subtraction. Based on the voice activation detection algorithm is studied and an improved spectrum subtraction( ISS) algorithm was presented. Our simulation results show that the proposed ISS Algorithm is effective with the lower computational complexity in speech enhancement system.
2 citations
01 Dec 2013
TL;DR: The results of simulation experiments indicate that the recognition accuracy of multi-band spectral subtraction robust speech recognition system is obviously superior to the basic spectral subtracted in different signal-to-noise ratios and different noise's types.
Abstract: In order to reduce the degradation of the speech recognition accuracy while the testing condition are mismatched with the training condition around noisy environment, a kind of multi-band spectral subtraction has been proposed. The estimated noise signals were extracted from the first few frames of the noisy speech. The noisy speech and estimation of noise signals by the frequency were divided into non-overlapping M frequency bands. According to the SNR (signal-to-noise ratio) of noise speech in each frequency band, the band noise spectral subtraction parameters can be determined. The front-end speech enhancement module and the speech recognizer constitute a robust speech recognition system. The results of simulation experiments indicate that the recognition accuracy of multi-band spectral subtraction robust speech recognition system is obviously superior to the basic spectral subtraction in different signal-to-noise ratios and different noise's types.
1 citations
Cites methods from "Robust Speech Recognition Based on ..."
...The other is the vector Taylor Series[4] and Cepstral Histogram Equalization[5] ....
[...]
...Vector Taylor Series in the logarithmic spectrum domain....
[...]
...The other is the vector Taylor Series[4] and Cepstral Histogram Equalization[5] .etc Feature Compensation Method....
[...]