Speech detection method for voice recognition system, involves calculating parameter representative of frame unit at group of fixed frame corresponding to speech frame with respect to another group of frame corresponding to noise frame

摘要

The method involves calculating a parameter representative of unit of a frame at a group of fixed frame corresponding to a speech frame with respect to another group of frame corresponding to noise frame. The parameter is analyzed to detect a nature of audio signal frame. Each group is represented by a Gaussian mixture representing probability density of each group. An independent claim is also included for a device of detection of speech of audio signal.