Design and Implementation of Emotion Recognition Algorithms on Spoken Sentences

Nowadays, there are many customer services for dealing with customers’ problem in company, like bank, 3C store. Sometime those calling people who pick up by rookie customer service are irrational or having difficult problem. When the customer can get a satisfied response, they will feel that the company doesn’t pay enough attention to them. So the purpose of this project is that design a real-time speech-recognition system. To achieve this goal, we split into two part and plan two year to execute. The first year, we will focus on create emotion databases, including emotion-speech-text corpus and acoustic database. Multi-media, like youtube, are the source of acoustic database; Social-media platform, like facebook, are the source of emotion speech-text corpus. For keeping freshness of databases, the system will training and learning continuously. The second year, we will design an emotion-voice-speech recognition algorithm based on the result of the first year. Not only extracting and adding the feature of voice, like tone and speed of sound into emotion-voice-speech recognition, combination of emotion-speech-text recognition result from emotion databases are the effective factors for precisely identifying voice emotions. In our research project result, we expect that we can provide many contributions in emotion speech-recognition research area.

Project ID:PB10703-1486
External Project ID:MOST106-2221-E182-076