Intelligent Robot is part of important applications in the area of artificial intelligence. As far as family nursing intelligent robot, apart from its computing ability and hardware requirement, it will also seek for the convenient, fast and user-friendly man-machine interactive mode, especially for the aged. Endpoint detection, isolated word and Special sound recognition and sound source location technologies will be wanted in this family situation where it judges and locates the people who need salvation according to sounds. Isolated word and special sound recognition uses MFCC and LPCC feature extraction and GMM model to classify and recognize. And sound source location uses microphone array to collect speech signals, and to analyze as well as process them using the processing technique of digital signals, in order to ensure and track the location of the acoustic. This article adds a new feature names Peak/Area ratio to MFCC as well as improves the endpoint detection, at the same time, optimizes GMM and TDE algorithm, and according to the environment of family application, designs and brings out a rescue system applied to intelligent robot in the environment of family. Tests in real environment shows that the system has well feasibility and recognition performance. |