Neutral to Joyous Happy Emotion Conversion

摘要

A method to convert neutrally synthesised speech to joyous happy emotion by prosody modification and laughter synthesis is presented. Modified zero frequency filtering (ZFF) method is proposed to extract the excitation source features from dynamically varying happy emotive, synthesised speech and laughter sequence. Prosody source features of verbal (spoken), source and system features of non-verbal (laughter) segments of elicited happy emotion are analysed and incorporated in the neutral speech. Pitch contours of verbal segments of the happy emotion are divided into three components and analysed separately for incorporation. Laughter sequence is synthesised by modifying the Line spectral frequencies (LSF), pitch period and strength of excitation (SoE) contours of a synthesised vowel utterance. The effectiveness of the synthesized happy emotive speech is confirmed by the subjective evaluation.