Prosodic parameters in emotional speech

要旨

Importance of speech prosody is on the increase as spontaneous interaction between human and machines is asked for. This paper examines how prosody contributes emotions to speech. Major elements which determine the emotion are pitch, tempo, and stress of speech. The last two elements correspond to duration and power of syllables, respectively. We choose ve emotions to be tested; anger, surprise, sorrow, hate, and joy. To verify our analysis, we have implemented a speech synthesis module which can easily control prosodic parameters of output speech. Responses to the synthesized speech show that the parameters of anger, sorrow and hate are con rmed over 85 %. Experiment results also suggest that surprise and joy feelings tend to depend less on its prosody. 1 Elements of Prosody We break \prosody down into three parts; accent, rhythm and intonation. When a word is pronounced, we can hear some syllable sound more noticeable. We de ne the term \accent as what makes some syllable sound conspicuously in a word. Accent has three factors; stress, pitch and duration. Essentially, English accent is a bundle of all these factors. That is to say, in English, an accented syllable is pronounced louder, more high-pitched and longer than non-accented syllables. In Japanese, on the other hand, pitch is the only important factor in terms of accent (Table 1), i.e. an accented syllable is pronounced more high-pitched than other syllables. Table 1: English accent vs. Japanese accent English Japanese Stress + Pitch + + Duration + We de ne the term \rhythm as a passage of time of those stress, pitch and duration. Generally, English rhythm is called stress-timed rhythm and Japanese rhythm is called syllable-timed rhythm. That means that, in Japanese, every syllable is pronounced in almost the same duration. And this forms a Japanese rhythm. \Intonation is an overall pitch ow of an utterance. While accent and rhythm have tendency to decide characteristics of a language, intonation seems common to every language. We may say that intonation re ects the universal mental make-up of human being.

主題