keyword spotting and utterance verification is widely used in dialog systems. The capability to deal with words out of a vocabulary is an important factor to evaluate a dialog system. This paper presents the dynamic g...
详细信息
keyword spotting and utterance verification is widely used in dialog systems. The capability to deal with words out of a vocabulary is an important factor to evaluate a dialog system. This paper presents the dynamic garbage evaluation approach, which evaluates the confidence of the utterance in speech recognition with strong flexibility. The combination of the dynamic garbage evaluation approach and antikeyword models solves the problem of lack of flexibility of the explicit garbage models and the verification capability of on-line garbage modeling.
Due to the excellent ability of dynamically adjusting the observation scope when the analyzing frequency changed, wavelet transform has been successfully applied to the processing of a non-stationary speech signal. We...
详细信息
Due to the excellent ability of dynamically adjusting the observation scope when the analyzing frequency changed, wavelet transform has been successfully applied to the processing of a non-stationary speech signal. We use the wavelet transform (WT) result and combine it with the traditional CEP vector (CEP-WT) to produce a new feature as the front-end output of an auto speech recognition system. Evaluation and experiments of the new feature show that the proposed method can improve the recognition system performance promisingly.
作者:
Tranzai LeeFang ZhengWenhu WuCenter for Speech Technology
State Key Laboratory of Intelligent Technology and Systems Department of Computer Science and Technology and Systems Department of Computer Science and Technology Tsinghua University Beijing China
The variations of speakers' vocal tract shapes result in the variations of the formant positions and sequentially in the variances of the features extracted from every frame of speech. In order to remove or reduce...
详细信息
The variations of speakers' vocal tract shapes result in the variations of the formant positions and sequentially in the variances of the features extracted from every frame of speech. In order to remove or reduce the variations of the formant positions, a speaker adaptation method is proposed and investigated in this paper which is based on a frequency warp function (f.w.f.). The f.w.f. warps the frequency axis so that the variations can be reduced. For a given speaker, some frequency reference points are selected to help to get this f.w.f. by finding the relationship between the positions of these reference points before and after the warping. According to the new positions of those reference points for the given speaker, the f.w.f. can then be constructed. The experimental results show that this method reduces the error rate by an average of 14.5%.
keywords spotting and utterance verification is widely used in dialog *** capability to deal with the words out of vocabulary is an important factor to evaluate a dialog *** paper presents dynamic garbage evaluation a...
详细信息
keywords spotting and utterance verification is widely used in dialog *** capability to deal with the words out of vocabulary is an important factor to evaluate a dialog *** paper presents dynamic garbage evaluation approach,which evaluates the confidence of the utterance in speech recognition with strong *** combination of dynamic garbage evaluation approach and antikeyword models solves the problem of lack flexibility of the explicit garbage models and the verification capability of on-line garbage modeling.
A stable discrete time tracking control approach based on dynamic inversion using dynamic neural networks (DNNs) is developed in the paper for robotic manipulators with unknown dynamics nonlinearities. Two novel desig...
详细信息
A stable discrete time tracking control approach based on dynamic inversion using dynamic neural networks (DNNs) is developed in the paper for robotic manipulators with unknown dynamics nonlinearities. Two novel design technologies-dynamic inversion constructed by DNNs and the NN variable structure control-are used for adaptive tracking controller design. The robot control law is composed of the dynamic inversion of the DNN system, adaptive compensation and the NN variable structure control (VSC) components. The developed control scheme guarantees the global stability and tracking error convergence of the NN control system. Finally, the control performance of the proposed control approach is illustrated through the comparison studies with robot tracking control approach using static NNs.
In this paper, a geometrical approach for building neural networks is proposed. With the proposed approach, it is very easy to construct an efficient neural classifier to solve the handwritten Chinese character recogn...
详细信息
ISBN:
(纸本)0769507506
In this paper, a geometrical approach for building neural networks is proposed. With the proposed approach, it is very easy to construct an efficient neural classifier to solve the handwritten Chinese character recognition problem, as well as other pattern recognition problems of large scale. Experiments are conducted to evaluate the performance of the proposed approach and results obtained are promising.
In the standard of the third generation digital mobile communications,there are pilot symbols in each slot,namely training *** traditional methods estimate pilot channel using pilot information,estimate data channel b...
详细信息
In the standard of the third generation digital mobile communications,there are pilot symbols in each slot,namely training *** traditional methods estimate pilot channel using pilot information,estimate data channel by interpolation or decision-feedback,and then obtain transmitted information by channel *** Kalman filter is a commonly used methods for tracking channel *** this paper,we propose an algorithm which uses Kalman filter based on AR model to track time-varying *** investigate how to estimate channel and implement equalizer by Kalman filter in DS-CDMA *** apply two methods to obtain the AR coefficients,one is adaptive LMS algorithm,and the other is Durbin's recursion method. Simulations show algorithm is effective for the estimation of the frequence-selective fading channel.
Along with the analysis of color features in the hue, saturation and value (HSV) space, a new dividing method to quantize the color space into 36 non-uniform bins is introduced in this paper. Based on this quantizatio...
详细信息
The extended associative memory (AM) neural network (EAMNN) has the advantage of performing the classification in noisy environments. We propose a faster robust learning algorithm of EAMNN and a new error cost functio...
详细信息
The extended associative memory (AM) neural network (EAMNN) has the advantage of performing the classification in noisy environments. We propose a faster robust learning algorithm of EAMNN and a new error cost function based on weighted sum of standard output error and Hamming distance of output error, and the additional derivatives term of first hidden layer neural activation functions. The fast backpropagation training is based on a modified steepest descent method derived by changing the error function to update weights according to output error, thus it speeds up significantly training speed of the MLP and BAM. The algorithm can force the hidden-layer activation to be saturated to reduce sensitivity of the output values to input variables effectively. It improves robustness on classification performance, increases associative memory ability and accelerates training speed of EAMNN. The experiments verify that it is more powerful than other networks. Then we proposed a two level tree structure modular EAMNN for large-set pattern classification.
Along with the analysis of color features in the hue, saturation and value (HSV) space, a new dividing method to quantize the color space into 36 non-uniform bins is introduced in this paper. Based on this quantizatio...
详细信息
Along with the analysis of color features in the hue, saturation and value (HSV) space, a new dividing method to quantize the color space into 36 non-uniform bins is introduced in this paper. Based on this quantization method we propose a color-spatial method to include several spatial features of the colors in an image for retrieval. These features are area and position, which mean the zero-order and the first-order moments, respectively. Experiments on an image database of 838 images show that the algorithm performs well in precision and adaptability.
暂无评论