Hidden Neural Network for Complex Pattern Recognition: A Comparison Study with Multi- Neural Network Based Approach

摘要: When the feature space undergoes changes, owing to different operating and environmental conditions, multi-aspect classification is almost a necessity in order maintain performance of pattern recognition system improve robustness reliability decision making. This an important issue being investigated ANN research, many cases, problems can be solved more effectively by combining one or two other techniques rather than implementing exclusively. New learning methods, especially multiple classifier systems, are now actively studied applied recognition. So, main goal this paper propose hybrid models compare your complex problem: speech biomedical diagnosis. compare, obtained with (1) Multi-network RBF/LVQ structure, we use involves Learning Vector Quantization (LVQ) as competitive processor Radial Basis Function (RBF) classifier. (2) Hybrid HMM/MLP model using Multi Layer-Perceptron (MLP) estimate Hidden Markov Models (HMM) emission probabilities. For pre- classification, k-means clustering algorithm proposed obtain optimum information for training data models.