使用2D紋理特徵之語音情緒辨識系統

Speech Emotion Recognition using 2D texture features

王俊明(Chun-Ming Wang)

指導教授 ： 秦群立

中山醫學大學/健康管理學院/應用資訊科學學系碩士班/碩士(2013年)

摘要

由於現今人們工作壓力造成情緒產生許多負面影響，因此我們提出了一套語音情緒辨識系統使用2D的紋理特徵。首先，我們利用聲譜圖轉換的技術將一維的訊號轉成二維的訊號，接著我們使用三次曲線調整方法去調整語音的二維訊號，最後我們利用紋理特徵抽取法去抽取這個二維的訊號特徵，這些特徵會將利用類神經網路去作辨識。在實驗結果中，我們用了三個情緒資料庫去測試我們的系統，期便是成功率在74.2%到80.6%。

關鍵字

語音情緒辨識 ； 二維 ； 聲譜圖 ； 三次曲線調整方法 ； 紋理特徵 ； 類神經網路

並列摘要

Now, many people have series work stress, it will result in negative emotion. Therefore, this paper proposes Speech Emotion Recognition using 2D texture features. First, we use spectrogram technology to transform 1D signal into 2D signal. Next, we use cubic curve contrast enhancement method to enhance the 2D signal contrast. Final, the image texture extraction method will be used to extract the texture features of the 2D signal. In the experiment result, we use three emotional databases to test our proposed system and use artificial neural network to be our classifier. The system recognition successful rate is distributed in 74.2% to 80.6%.

並列關鍵字

Speech Emotion Recognition ； 2D Spectrogram ； Cubic Curve Contrast Enhancement ； Texture Feature ； Artificial Neural Network

參考文獻

[22] Chin-Teng Lin and Chiun-Li Chin, “Using Fuzzy Inference and Cubic Curve to Detect and Compensate Backlight Image,” International Journal of Fuzzy Systems, vol. 8, no. 1, March 2006.

[1] R. E. Thayer, The Biopsychology of Mood and Arousal, 1989.

[3] Calvo, R. A., & D’Mello, and S. “Affect detection: An interdisciplinary review of models, methods, and their applications.” IEEE Transactions on Affective Computing, vol.1, 18–37, 2011.

[4] I. Shahin, “Speaker recognition systems in the emotional environment,” in Proc. IEEE ICTTA ’08, Damascus, Syria, 2008.

[5] M. V. Ghiurcau, C. Rusu, and J. Astola, “A study of the effect of emotional state upon text-independent speaker identification,” in Proc. ICASSP 2011, Prague, Czech Republic, 2011.