研究了基于卷积神经网络的语音情感识别算法，改进了传统卷积神经网络训练过程中的卷积核权值的更新算法，使卷积核权值的更新算法与迭代次数有关联；同时为了增加情感语音之间的特征差异性，将语音信号经过预处理后得到的梅尔频率倒谱系数特征数据矩阵进行变换，提高卷积神经网络的表达能力.实验表明，改进后的语音情感识别算法的错误识别率比传统算法的错误识别率约减少7%.
In this paper, we studied the algorithm of speech emotion recognition based on convolutional neural networks, and improved the algorithm of updating convolution kernel weight during the training process of traditional convolutional neural networks, resulting that the algorithm of updating the convolution kernel weight was related to the number of iterations. Simultaneously, in order to increase the difference of emotional phonetic features, the data matrix of the Mel-frequency cepstral coefficients (MFCC) obtained by preprocessing the speech signal was transformed, consequently, improved the expressive ability of convolutional neural networks. Experiments showed that the error recognition rate of the improved algorithm of speech emotion recognition was about 7% lower than that of traditional algorithms.
[1] Anagnostopoulos C N, Iliou T, Giannoukos I. Features and classifiers for emotion recognition from speech:a survey from 2000 to 2011[J]. Artificial Intelligence Review, 2015, 43(2):155-177.
[2] Juang B H, Rabiner L. Mixture autoregressive hidden Markov models for speech signals[J]. Procedia Computer Science, 2015, 61(6):328-333.
[3] Vlassis N, Likas A. A greedy EM algorithm for Gaussian mixture learning[J]. Neural Processing Letters, 2002, 15(1):77-87.
[4] Hu H, Xu M X, Wu W. GMM supervector based SVM with spectral features for speech emotion recognition[C]//IEEE International Conference on Acoustics, 2007:IV-413-IV-416.
[5] Lee C M, Yildirim S, Bulut M, Kazemzadeh A, Busso C, Deng Z, Lee S, Narayanan S. Emotion recognition based on phoneme classes[J]. Proc. icslp Oct, 2004:889-892.
[6] Mao Q, Dong M, Huang Z, Zhan Y. Learning salient features for speech emotion recognition using convolutional neural networks[J]. IEEE Transactions on Multimedia, 2014, 16(8):2203-2213.
[7] Zhang B, Quan C, Ren F. Performance of convolution neural network on the recognition of speech emotion and images[C]//AIA International Advanced Information Institute, 2016:12-21.
[8] Zheng W Q, Yu J S, Zou Y X. An experimental study of speech emotion recognition based on deep convolutional neural networks[C]//International Conference on Affective Computing and Intelligent Interaction, 2015:827-831.
[9] 郭鹏娟. 语音情感特征提取方法和情感识别研究[D]. 西安:西北工业大学,2007.
[10] Davis S, Mermelstein P. Comparison of parametric representations for monosyllabic word recognition in continuously spoken sentences[J]. Readings in Speech Recognition, 1990, 28(4):65-74.
[11] Hinton G, Deng L, Yu D, Dahl G E, Mohamed A, Jaitly N, Senior A, Vanhoucke V, Nguyen P, Sainath T N, Kingsbury B. Deep neural networks for acoustic modeling in speech recognition:the shared views of four research groups[J]. IEEE Signal Processing Magazine, 2012, 29(6):82-97.
[12] Lecun Y. Convolutional networks for images, speech, and time-series[J]. Handbook of Brain Theory & Neural Networks, 1995.
[13] Krizhevsky A, Sutskever I, Hinton G E. Imagenet classification with deep convolutional neural networks[C]//International Conference on Neural Information Processing Systems, 2012:1097-1105.
[14] Lecun Y, Bottou L, Bengio Y, Haffner P. Gradient-based learning applied to document recognition[J]. Proceedings of the IEEE, 1998, 86(11):2278-2324.
