To achieve efficient a d compact low-dimensional features for speech emotion recognition,a novel featurereduction method using uncertain linear discriminant analysis is proposed.Using the same principles as for conven...To achieve efficient a d compact low-dimensional features for speech emotion recognition,a novel featurereduction method using uncertain linear discriminant analysis is proposed.Using the same principles as for conventional linear discriminant analysis(LDA),uncertainties of the noisy or distorted input data ae employed in order to estimate maximaiy discriminant directions.The effectiveness of the proposed uncertain LDA(ULDA)is demonstrated in the Uyghur speech emotion recognition task.The emotional features of Uyghur speech,especially,the fundamental fequency and formant,a e analyzed in the collected emotional data.Then,ULDA is employed in dimensionality reduction of emotional features and better performance is achieved compared with other dimensionality reduction techniques.The speech emotion recognition of Uyghur is implemented by feeding the low-dimensional data to support vector machine(SVM)based on the proposed ULDA.The experimental results show that when employing a appropriate uncertainty estimation algorithm,uncertain LDA outperforms the conveetional LDA counterpart on Uyghur speech emotion recognition.展开更多
Some two-microphone noise reduction techniques that work in the frequency domain exploit coherence function between two noisy signals. They have shown good results when noise signals on two sensors are uncorrelated, b...Some two-microphone noise reduction techniques that work in the frequency domain exploit coherence function between two noisy signals. They have shown good results when noise signals on two sensors are uncorrelated, but their per-formance decreases with correlated noises. Coherence based methods can be improved when the cross power spectral density (CPSD) of correlated noise signals is available. In this paper, we propose a new method for estimation of the CPSD of the noise, which is based on the minimum tracking technique. Despite the fact that the proposed estimator does not need to implement a voice activity detector (VAD), its performance is comparable to a CPSD estimator that uses an ideal VAD.展开更多
基金The National Natural Science Foundation of China(No.61673108,61231002)
文摘To achieve efficient a d compact low-dimensional features for speech emotion recognition,a novel featurereduction method using uncertain linear discriminant analysis is proposed.Using the same principles as for conventional linear discriminant analysis(LDA),uncertainties of the noisy or distorted input data ae employed in order to estimate maximaiy discriminant directions.The effectiveness of the proposed uncertain LDA(ULDA)is demonstrated in the Uyghur speech emotion recognition task.The emotional features of Uyghur speech,especially,the fundamental fequency and formant,a e analyzed in the collected emotional data.Then,ULDA is employed in dimensionality reduction of emotional features and better performance is achieved compared with other dimensionality reduction techniques.The speech emotion recognition of Uyghur is implemented by feeding the low-dimensional data to support vector machine(SVM)based on the proposed ULDA.The experimental results show that when employing a appropriate uncertainty estimation algorithm,uncertain LDA outperforms the conveetional LDA counterpart on Uyghur speech emotion recognition.
基金Project supported by the Iran Telecommunications Research Center (ITRC)
文摘Some two-microphone noise reduction techniques that work in the frequency domain exploit coherence function between two noisy signals. They have shown good results when noise signals on two sensors are uncorrelated, but their per-formance decreases with correlated noises. Coherence based methods can be improved when the cross power spectral density (CPSD) of correlated noise signals is available. In this paper, we propose a new method for estimation of the CPSD of the noise, which is based on the minimum tracking technique. Despite the fact that the proposed estimator does not need to implement a voice activity detector (VAD), its performance is comparable to a CPSD estimator that uses an ideal VAD.
