Detecting Emotions in Mandarin Speech

Detecting Emotions in Mandarin Speech Tsang-Long Pao, Yu-Te Chen, Jun-Heng Yeh and Jhih-Jheng Lu Department of Computer Science and Engineering, Tatung University, Taipei tlpao@ttu.edu.tw, d8906005@mail.ttu.edu.tw, g9106004@ms2.ttu.edu.tw Abstract. In this paper, a Mandarin speech based emotion classification method is presented. Five primary human emotions including anger, boredom, happiness, neutral and sadness are investigated. For speech emotion recognition, we select 16 LPC coefficients, 12 LPCC components, 16 LFPC components, 16 PLP coefficients, 20 MFCC components and jitter as the basic features to form the feature vector. Two text-dependent and speaker-independent corpora are employed. The recognizer presented in this paper is based on three recognition techniques: LDA, K-NN, and HMMs. Results show that the selected features are robust and effective in the emotion recognition at the valence degree in both corpora. For the LDA emotion recognition, the highest accuracy of 79.9% is obtained. For the K-NN emotion recognition, the highest accuracy of 84.2% is obtained. And for the HMMs emotion recognition, the highest accuracy of 88.7% is achieved. 1 Introduction Various opinions of emotions proposed by more than 100 scholars are summarized in a classical article [1]. Research on the cognitive component focuses on understanding the environmental and attended situations that gives rise to emotions; research on the physical components emphasizes the physiological response that co-occurs with an emotion or rapidly follows it. In short, emotions can be considered as communications to oneself and others [1]. Emotions consist of behaviors, physiologic changes and subjective experience as evoked by individual’s thoughts, socio-cultures and so on. Emotions are traditionally classified into two main categories: primary (basic) and secondary (derived) emotions [2]. Primary or basic emotions generally could be experienced by all social mammals (e.g. humans, monkeys, dogs, whales) and have particular manifestations associated with them (e.g. vocal/facial expressions, behavioral tendencies, and physiological patterns). Secondary or derived emotions are the combination or derivation from primary emotions. Emotional dimensionality is a simplified description of basic properties of emotional states. According to Osgood, Suci and Tannenbaum’s theory [3] and subsequent psychological research [4], the computing of emotions is conceptualized as three major dimensions of connotative meaning, arousal, valence and power. In general, the arousal and valence dimensions can be used to distinguish most basic emotions. The emotions location in arousal-valence space is shown in Fig. 1 [3], which results in a representation that is both simple and capable of conforming to wide emotional applications. Fig. 1. Graphic representation of the arousal-valence theory of emotions positive negative valence low high anger happiness sadness boredom neutral arousal Table 1. Emotions and speech relations Anger Happiness Sadness Fear Disgust Speech Rate Slightly faster Faster or slower Slightly slower Much faster Very much faster Pitch Average Very much higher Much higher Slightly lower Very much higher Very much lower Pitch Range Much wider Much wider Slightly narrower Much wider Slightly wider Intensity Higher Higher Lower Normal Lower Voice Quality Breathy, chest Breathy, blaring tone Resonant Irregular voicing Grumble chest tone Pitch changes Abrupt on stressed Smooth, upward inflections Downward inflections Normal Wide, …

Content truncated to 500 words