Hybrid Feature Embedded Sparse Stacked Autoencoder and Manifold Dimensionality Reduction Ensemble for Mental Health Speech Recognition

Speech feature learning is the key to speech mental health recognition. Deep can automatically extract features but suffers from small sample problem. The traditional method effective, cannot find inter-feature structure generate new high-quality features. This paper proposes an embedded hybrid deep sparse stacked autoencoder ensemble solve this Firstly, are extracted based on prior knowledge and called original Secondly, into network (Sparse Stacked Autoencoder) filter output of hidden layer, enhance complementarity between Thirdly, L1 regularized selection mechanism designed reduce set formed by combination Finally, a manifold projection classifier stability classification. Besides, for first time collection scheme We construct large-scale Chinese database verification proposed algorithm health. In experimental section, verified compared with representative related algorithms. results show that has better classification accuracy than other combines advantages extraction methods more efficiently