Method for identity authentication and reliable control based on speech recognition

The present invention discloses a method for identity authentication and reliable control based on speech recognition. The method comprises speech signal preprocessing, feature extraction modeling, training of a template library and mode matching. The speech signal preprocessing comprises the steps of: performing windowing processing of collected audio signals, performing end point detection of the audio signals, and allowing a computer to find out start points and end points of characters and words from signals comprising a section of speech; and performing short-time average zero-crossing rate calculation, distinguishing unvoiced sound volumes and voiced sound volumes, and distinguishing sound and silence in a condition with a high signal to noise ratio. The feature extraction modeling comprises the steps of: extracting energy, amplitude, a frequency spectrum, a zero-crossing rate and a power spectrum of collected speech as feature parameters, and filtering redundant information. Themethod provided by the i