Research on home environment recognition system based on Gaussion Mixture Model

This paper presents the design and implementation of a new recognition method to monitor home accident and anomalies based on home environmental voice. Gaussian Mixture Models theory is utilized to modeling and identifies home environmental situations, so that it can perform real time detection and recognition for solitaries. The whole process includes four steps: pretreatment of voice signal, extracting characteristic parameters, training GMM model of home environmental condition, recognition and alarm output. The experiment shows effectiveness of the designed method.