Emotion recognition and synthesis system on speech

Abstract

A system that is capable of both recognizing and synthesising emotional content in speech is developed. First, the information that relates the physical features of emotional speech to the emotional content perceived by the listeners is estimated through linear statistical methods and it is applied to the system. The system realises emotion recognition and synthesis by means of a simple linear operation using the relation information. In the system, the pitch contour is expressed by the seven-parameter model proposed by Hirose, Fujisaki & Yamaguchi (1984), the power envelope is approximated by five line segments (11 parameters), and PSOLA (Pitch-Synchronous OverLAp) is applied to synthesise the speech. A set of emotional words, among which there is very little correlation, was selected from the preliminary statistical experiments. The relation information was verified as being significant and, from the results of the experiments, the system was able to recognise and synthesise emotional content in speech just as human subjects did.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 53 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.