Alpine Electronics, Inc.

Shingo Kiuchi

Nozomu Saito

Koichi Nakata

Kouichi Nakata

Toshiaki Asano

Speech recognition performance is improved without changing a speech recognition engine. A speech data generation section generates, from speech data for which speech recognition is to be performed, a plurality of pieces of speech data whose starting positions of the non-speech regions differ. A speech recognition engine performs speech recognition by using each of the pieces of speech data. A totaling/comparison section provides the most numerous recognized result from among a plurality of obtained recognized results.

Speech recognition performance improvement method and speech recognition device

An input sound processor compares power at each frequency component of an input sound with a reference value, and sets multiplication points indicating frequency components at which the total power of the input sound is to be determined. A product-sum operation is performed at the multiplication points on the power at each frequency component and the square amplitude of each filter coefficient indicating the transfer characteristic from a loudspeaker to a microphone to estimate the total power of the input sound at the position of the microphone.

Input sound processor

The microphone system of the invention executes an adaptive filter processing by using output signals from two microphones to output a speaker's voice signal with an improved SN ratio, in which the two microphones are laid out close to each other, and the angles formed by the orientations of the microphones with respect to the speaker's vocalizing direction are made different for each of the microphones. For example, the microphones are mounted on the sun visor of a vehicle, or on the ceiling above the front passenger seat or the driver's seat of the vehicle, with the orientations of the microphones differentiated. Further, the SN ratio of the output signal from one microphone is raised, and the SN ratio of the output signal from the other microphone is lowered. For example, one microphone is positioned right above a speaker's face, and the other microphone is spaced apart on the occipital side by about 1 to 5 cm from the position of the first microphone. Thus, the microphone system improves the SN ratio of the voice signal.

Microphone system

In a conventional device for extracting voice features accurately without being influenced by noises, such as a voice recognition device, usually an input voice signal is processed first by a noise reduction system having the tap length N, and the result is FFT-processed by L-points, and then the power spectrum vector is calculated; accordingly, a one time operation requires N multiplications and (N&#x2212;1) summations. The voice feature extraction device according to the invention receives a voice signal including noises from a microphone, which is processed by a window function operation unit, and thereafter FFT-processed by an FFT operation unit by L-points. A power calculation unit calculates a power spectrum vector of the input voice signal. However, a noise reduction system determines in advance a filter coefficient of this system and processes the coefficient to calculate a noise reduction coefficient, and the power spectrum vector is processed by this noise reduction system. Thereby, the voice feature extraction device of the invention reduces the processing volume to 1/(4N&#x2212;2) in comparison to the conventional device, lightens the processing load of the processing unit, and increases the processing speed.

Voice feature extraction device

The present invention provides a voice input apparatus which can extract an input voice without interrupting a generated background sound. When a talk switchis pushed down, a LMS algorithm processing partstops an update operation of a filter coefficient of an adaptive filterand stores filter coefficient W1 at this moment in a portion for storing filter coefficients. A portion for calculating filter coefficientscalculates a filter coefficient W2 of a filterto simulate a sound-transmittal line of an audio sound which is produced from a speakerand is input into microphonesby means of these stored filter coefficient W1 and various transmittal characteristics CS1, CS2 and the like. By extracting an output signal Yof the filterfrom an output signal Yof a portion for eliminating surrounding noiseby means of a computing part, audio sound components included with an operating voice collected by the microphonesare removed.

Growing community of inventors

Iwaki, Japan

Shingo Kiuchi