Based Maritime Surveillance System, Voice Recognition Technology Research

Wireless sensor has a broad application in modern military. The contour and sound of military objective can identify by military monitoring system on the sea with sensors instead of man. So many wireless sensors are needed that a system which simple, cheap and reliable sensor system to have a tentative perception of sound is in need.Own to the complexity of the sound of ships, planes, seabirds. passive sonar and marine background, recognition of these sounds is different from speech recognition in which many traditional feature extraction techniques and recognition technologies is used. The usability of these technologies will be investigated.The study focuses on environmental sounds recognition after the monitoring center receiving audio data from the sea. First a brief introduction is given to sound propagation and detection process. Then the suitability of several environmental sound recognition techniques is tested based on a comprehensive investigation of technologies on speech recognition, non-speech environmental sound classification, musical instrument recognition and passive sonar classification. After that, two recognition system is implemented:one is implementation of sound recognition using dynamic time warping technique combined with Fourier transform. Mel-cepstral coefficients and the short time Fourier transform feature extraction techniques under MATLAB7.0; the other is building Gaussian Mixture Model using the Hidden Markov Model Toolkit. The performance of each system is evaluated by recognition rate and calculating time. Besides, functional recognition software with a friendly interface using fast Fourier transform and dynamic time warping algorithm is developed.