| 摘要 |
PURPOSE:To extract a feature of a voice which is not varied due to an environment and a condition, and also, to contrive the improvement of a recognition rate of a voice recognizing device by constituting the device so that a parameter in a half-wave rectifying part for showing a response characteristic of an auditory sense is not fixed but determined in accordance with an output level of a preprocessing part. CONSTITUTION:An output of every channel of a critical band pass filter 20 is inputted to an automatic gain adjusting part 30, and from an output of the automatic gain adjusting part 30, a degree of saturation of a half-wave rectification is calculated in a saturation degree calculating part 40. Subsequently, an output of every channel of the automatic gain adjusting part 30, and the degree of saturation of the half-wave rectification are inputted to a half-wave rectifying part 50, and an output of the half-wave rectifying part 50 becomes an output of an auditory model analyzing part 7. In such a way, a satisfactory feature parameter is obtained, and a recognition rate of a voice recognizing device can be improved. |
