Speech emotion recognition based on an improved supervised manifold learning algorithm

To improve effectively the performance on speech emotion recognition, it is needed to perform nonlinear dimensionality reduction for speech feature data lying on a nonlinear manifold embedded in high-dimensional acoustic space. Supervised Locally Linear Embedding (SLLE) is a typical supervised manifold learning algorithm for nonlinear dimensionality reduction. Considering the existing drawbacks of SLLE, this paper proposes an improved version of SLLE, which enhances the discriminating power of low-dimensional embedded data and possesses the optimal generalization ability. The proposed algorithm is used to conduct nonlinear dimensionality reduction for 48-dimensional speech emotional feature data including prosody and voice quality features, and extract low-dimensional embedded discriminating features so as to recognize four emotions including anger, joy, sadness and neutral. Experimental results on the natural speech emotional database demonstrate that the proposed algorithm obtains the highest accuracy of 90.78% with only less 9 embedded features, making 15.65% improvement over SLLE. Therefore, the proposed algorithm can significantly improve speech emotion recognition results when applied for reducing dimensionality of speech emotional feature data.