In conventional source-filter models, voiced and unvoiced components were considered independently. However, in practice it was difficult to separate the source into two parts. An actual source consists of a mixture o...In conventional source-filter models, voiced and unvoiced components were considered independently. However, in practice it was difficult to separate the source into two parts. An actual source consists of a mixture of two sources and the ratio varies according to the content or the intention of speaker. It had been investigated to separate the voiced and unvoiced components for different source models. Source signals were modeled based on the residual signal measured from inverse filtering. Three different source models were assumed. The parameters of each model were optimized for the original speech signal using a genetic algorithm. The resulting parameters were compared in terms of the mel-cepstral distance to the original signal, the spectrogram and the spectral envelope from the synthesized signal. The optimization method achieves an improvement of 15% for the Klatt model, but there is little improvement in the modified residual case.展开更多
This work is concerned with the development and optimization of a signal model for scalable perceptual audio coding at low bit rates. A complementary two-part signal model consisting of Sines plus Noise (SN) is descri...This work is concerned with the development and optimization of a signal model for scalable perceptual audio coding at low bit rates. A complementary two-part signal model consisting of Sines plus Noise (SN) is described. The paper presents essentially a fundamental enhancement to the sinusoidal modeling component. The enhancement involves an audio signal scheme based on carrying out overlap-add sinusoidal modeling at three successive time scales, large, medium, and small. The sinusoidal modeling is done in an analysis-by-synthesis overlap- add manner across the three scales by using a psychoacoustically weighted matching pursuits. The sinusoidal modeling residual at the first scale is passed to the smaller scales to allow for the modeling of various signal features at appropriate resolutions.This approach greatly helps to correct the pre-echo inherent in the sinusoidal model. This improves the perceptual audio quality upon our previous work of sinusoidal modeling while using tile same number of sinusoids. Tile most obvious application for the SN model is in scalable, high fidelity audio coding and signal modification.展开更多
This letter proposes an effective and robust speech feature extraction method based on statistical analysis of Pitch Frequency Distributions (PFD) for speaker identification. Compared with the conventional cepstrum, P...This letter proposes an effective and robust speech feature extraction method based on statistical analysis of Pitch Frequency Distributions (PFD) for speaker identification. Compared with the conventional cepstrum, PFD is relatively insensitive to Additive White Gaussian Noise (AWGN), but it does not show good performance for speaker identification, even if under clean environments. To compensate this shortcoming, PFD and conventional cepstrum are combined to make the ultimate decision, instead of simply taking one kind of features into account.Experimental results indicate that the hybrid approach can give outstanding improvement for text-independent speaker identification under noisy environments corrupted by AWGN.展开更多
This paper proposed improvements to the low bit rate parametric audio coder with sinusoid model as its kernel. Firstly, we propose a new method to effectively order and select the perceptually most important sinusoids...This paper proposed improvements to the low bit rate parametric audio coder with sinusoid model as its kernel. Firstly, we propose a new method to effectively order and select the perceptually most important sinusoids. The sinusoid which contributes most to the reduction of overall NMR is chosen. Combined with our improved parametric psychoacoustic model and advanced peak riddling techniques, the number of sinusoids required can be greatly reduced and the coding efficiency can be greatly enhanced. A lightweight version is also given to reduce the amount of computation with only little sacrifice of performance. Secondly, we propose two enhancement techniques for sinusoid synthesis: bandwidth enhancement and line enhancement. With little overhead, the effective bandwidth can be extended one more octave; the timbre tends to sound much brighter, thicker and more beautiful.展开更多
基金supported by the Second Stage of Brain Korea 21 Projects
文摘In conventional source-filter models, voiced and unvoiced components were considered independently. However, in practice it was difficult to separate the source into two parts. An actual source consists of a mixture of two sources and the ratio varies according to the content or the intention of speaker. It had been investigated to separate the voiced and unvoiced components for different source models. Source signals were modeled based on the residual signal measured from inverse filtering. Three different source models were assumed. The parameters of each model were optimized for the original speech signal using a genetic algorithm. The resulting parameters were compared in terms of the mel-cepstral distance to the original signal, the spectrogram and the spectral envelope from the synthesized signal. The optimization method achieves an improvement of 15% for the Klatt model, but there is little improvement in the modified residual case.
基金Supported by the National Natural Science Foundation of China(No.69802007)Motorola China Research Center(No.B38300)Natural Science Foundation of Guangdong(No.011611)
文摘This work is concerned with the development and optimization of a signal model for scalable perceptual audio coding at low bit rates. A complementary two-part signal model consisting of Sines plus Noise (SN) is described. The paper presents essentially a fundamental enhancement to the sinusoidal modeling component. The enhancement involves an audio signal scheme based on carrying out overlap-add sinusoidal modeling at three successive time scales, large, medium, and small. The sinusoidal modeling is done in an analysis-by-synthesis overlap- add manner across the three scales by using a psychoacoustically weighted matching pursuits. The sinusoidal modeling residual at the first scale is passed to the smaller scales to allow for the modeling of various signal features at appropriate resolutions.This approach greatly helps to correct the pre-echo inherent in the sinusoidal model. This improves the perceptual audio quality upon our previous work of sinusoidal modeling while using tile same number of sinusoids. Tile most obvious application for the SN model is in scalable, high fidelity audio coding and signal modification.
文摘This letter proposes an effective and robust speech feature extraction method based on statistical analysis of Pitch Frequency Distributions (PFD) for speaker identification. Compared with the conventional cepstrum, PFD is relatively insensitive to Additive White Gaussian Noise (AWGN), but it does not show good performance for speaker identification, even if under clean environments. To compensate this shortcoming, PFD and conventional cepstrum are combined to make the ultimate decision, instead of simply taking one kind of features into account.Experimental results indicate that the hybrid approach can give outstanding improvement for text-independent speaker identification under noisy environments corrupted by AWGN.
文摘This paper proposed improvements to the low bit rate parametric audio coder with sinusoid model as its kernel. Firstly, we propose a new method to effectively order and select the perceptually most important sinusoids. The sinusoid which contributes most to the reduction of overall NMR is chosen. Combined with our improved parametric psychoacoustic model and advanced peak riddling techniques, the number of sinusoids required can be greatly reduced and the coding efficiency can be greatly enhanced. A lightweight version is also given to reduce the amount of computation with only little sacrifice of performance. Secondly, we propose two enhancement techniques for sinusoid synthesis: bandwidth enhancement and line enhancement. With little overhead, the effective bandwidth can be extended one more octave; the timbre tends to sound much brighter, thicker and more beautiful.
