In the audio register method for speech recognition, the process which analyzes the spectrum of the correspondence number which is input from outside; The process which extracts the language unit which beforehand from the aural signal inside the aforementioned correspondence number is set for speaker recognition; The process which measures the voice of the aforementioned each language unit; As a standard the register which includes the voice data of the group of background speakers (the background) the process which collects the voice data of the speaker in the audio database; The process which is judged whether or not the voice of the aforementioned each language unit it is inside the voice range which is set beforehand on the basis of the aforementioned audio database; When it is inside the voice range where language units of a quantity which is set beforehand are beforehand set, using the multilayer nerve net, the process which studies each language unit; And the description above the process which houses the data of each language unit which was studied as the data in order to recognize the aforementioned speaker; It includes. Because of this, compared to it can verify the speaker accurately with when studying in order to register the sound of the speaker and considering the voice of the speaker when verifying the speaker.
展开▼