Kernel fuzzy c-means speech emotion identification method combined with secondary identification of support vector machine

Abstract

The invention relates to a kernel fuzzy c-means speech emotion identification method combined with the secondary identification of a support vector machine. The kernel fuzzy c-means speech emotion identification method comprises the following steps of: firstly, using a Mel-frequency cepstral coefficient to map to a high dimension characteristic space by a nonlinear kernel function; obtaining a clustering center by utilizing the kernel fuzzy c-means and using the clustering center as a code book quantized by a vector; then estimating a sample category label by an average fuzzy weighted minimum vector quantized error rule; and after estimating the label, carrying out secondary identification on pleasure and anger by utilizing a support vector machine method and a linear prediction coefficient. The kernel fuzzy c-means speech emotion identification method is an important composition part for automatically identifying the emotion in a man-machine interactive system, is a first step of carrying out emotional interaction between the machine and a person and has important application prospect for a system which uses people as the design center.

Description

Technical field [0001] The invention belongs to the cross field of speech processing and artificial intelligence, and particularly relates to an emotion recognition method in an intelligent human-computer interaction system. Background technique [0002] At present, speech emotion pattern recognition methods based on traditional statistical methods, or machine learning such as neural networks, these traditional classic methods can only achieve the ideal recognition rate when the sample data is large enough, so when the actual sample data When less or limited, the engineering application of traditional methods will be limited. In addition, it is well known that the expressions of natural emotions and the presented states are often more ambiguous or mixed (see literature: Emotion classification from speech using evaluator reliability-weighted combination of ranked-lists, Audhkhasi, K., Narayanan, S. , In ICASS, Issue: July, 2011, pp: 4956-4959.). Therefore, when designing a speec...

AI Technical Summary

Problems solved by technology

Therefore, the mixed attribute of emotion should be considered when designing a speech emotion recognition system, so the previous recognition method is not optimal only to give the unknown emotion a separate prediction label (see literature: A Framework for Automatic Human Emotion Classification Using Emotion Profiles, Emily M., Narayanan, S., IEEE Trans On Audio, Speech and Language Processing, 2011, vol.19, pp.1507-1520)

Images