Extracting Emotion from Voice

Authors: S. Fukuda, V. Kostov
Tokyo Metro. Institute of Technology

Emotions are gleaned mostly from voice, facial expressions, and body language. Humans have a fairly accurate ability to determine emotions from just voice. Humans distinguish four basic characteristics of sound: intensity, pitch, color, and direction. By analyzing Wavelets/Cepstrum the authors appear to have been able to get emotions from wav file inputs. The paper talks about the method they used, but isn't that specific about the accuracy of their method. They found that there is a difference between Japanese and English inputs, but weren't that specific about what the differences were. The paper concludes with saying they have taken a first step towards "Emotional Sensors", but that the field has a long way to go.

Quality: 3/5
Relevance: 5/5