Swarm Intelligence Based Feature Clustering for Continuous Speech Recognition Under Noisy Environments

Abstract

Swarm Intelligence based Feature Clustering using Artificial Bee Colony (ABC) technique is proposed and implemented in this research work. It is used to group and label the features of continuous speech sentence. This algorithm is unsupervised classification which classifies the feature vectors into different clusters and it will enhance the quality of clustering. Simulation is carried out for various number of clusters for different speech recognition algorithms under clean and noisy environmental conditions of NOIZEUS database signals. Experimental results reveal that the proposed hybrid clustering provides substantial enhancements in various performance measures compared with the existing algorithms and found as number of cluster as 5 produces the optimal result. For this optimal value, ABC clustering technique provides optimal enhancement in the recognition accuracy of Continuous Speech Recognition compared with the existing clustering techniques under different speech signal environments.