Abstract: A band extending apparatus (1) is provided with: first generating device (111, 112) for generating a baseband signal (XB(n)) by up-sampling an input signal (X(n)) and then transmitting it through a low-pass filter; a second generating device (21) for generating a high-frequency signal (XH(n)), by extracting a signal component on a higher-frequency side of a signal which is obtained by squaring a band limited signal (Xb(n)) which is a signal component with a predetermined band of the baseband signal; and a third generating device (141) for generating an output signal (XE(n)) by adding the high-frequency signal to the baseband signal.

Abstract: An operator recognition device is provided that eliminates the registration of data such as HMM data having a characteristic amount for which error in recognition occurs easily when recognizing an operator, and thus reduces the possibility of errors in recognition, and has stable recognition performance. When registering HMM data that is used when performing recognition processing, a speaker recognition device 100 eliminates the registration of HMM data of a password having a characteristic amount of the spoken voice component that is similar to a characteristic amount that is indicated by HMM data that is already registered, and does not allow the registration of HMM data for which it is estimated that error in recognition will occur easily during the recognition process.

Abstract: A noise suppression apparatus calculates a sound spectrum and a noise spectrum from an input sound, further calculates gain based on the sound spectrum and noise spectrum, and suppresses noise in the input sound. The noise suppression apparatus includes a first frame-dividing unit that divides the input sound into frames having a predetermined frame length, a second frame-dividing unit that divides the input sound into frames having a longer frame length than the frame length of the first frame-dividing unit, a second converting unit that converts, into a spectrum, the input sound divided into frames by the second frame-dividing unit, a smoothing unit that smoothes the converted spectrum in a frequency direction, and a gain calculating unit that calculates gain based on the smoothed spectrum and the noise spectrum.

Abstract: A band extending apparatus (1) is provided with: first generating device (111, 112) for generating a baseband signal (XB(n)) by up-sampling an input signal (X(n)) and then transmitting it through a low-pass filter; a second generating device (21) for generating a high-frequency signal (XH(n)), by extracting a signal component on a higher-frequency side of a signal which is obtained by squaring a band limited signal (Xb(n)) which is a signal component with a predetermined band of the baseband signal; and a third generating device (141) for generating an output signal (XE(n)) by adding the high-frequency signal to the baseband signal.

Abstract: An operator recognition device is provided that eliminates the registration of data such as HMM data having a characteristic amount for which error in recognition occurs easily when recognizing an operator, and thus reduces the possibility of errors in recognition, and has stable recognition performance. When registering HMM data that is used when performing recognition processing, a speaker recognition device 100 eliminates the registration of HMM data of a password having a characteristic amount of the spoken voice component that is similar to a characteristic amount that is indicated by HMM data that is already registered, and does not allow the registration of HMM data for which it is estimated that error in recognition will occur easily during the recognition process.

Abstract: A noise suppression apparatus calculates a sound spectrum and a noise spectrum from an input sound, further calculates gain based on the sound spectrum and noise spectrum, and suppresses noise in the input sound. The noise suppression apparatus includes a first frame-dividing unit that divides the input sound into frames having a predetermined frame length, a second frame-dividing unit that divides the input sound into frames having a longer frame length than the frame length of the first frame-dividing unit, a second converting unit that converts, into a spectrum, the input sound divided into frames by the second frame-dividing unit, a smoothing unit that smoothes the converted spectrum in a frequency direction, and a gain calculating unit that calculates gain based on the smoothed spectrum and the noise spectrum.

Abstract: A pulse width modulator for producing a PWM signal having reduced nonlinear distortion through interpolation processing with fewer computation steps is provided. The computational processing, W={1+?(X2?X1)}{0.25 (X1+X2)}+0.5, is performed in accordance with successive sample values (X1, X2) in a PCM data train to determine a weighting factor (W). The computational processing, Xq=0.5·X0(W2?W)+X1(1?W2)+0.5·X2(W2+W), is performed using the sample values (X1, X2), a sample value (X0) previous to the sample value (X1), and the weighting factor (W) to thereby determine an interpolated sample value (Xq) having an amplitude close to that of an original analog signal (X(t)) generating the PCM data train. A point in time (tq) at which a reference signal (R(t)) takes on the interpolated sample value (Xq) is then determined to produce a PWM signal (Spwm) which is logically inverted at the point in time (tq).

Abstract: An initial combination HMM 16 is generated from a voice HMM 10 having multiplicative distortions and an initial noise HMM of additive noise, and at the same time, a Jacobian matrix J is calculated by a Jacobian matrix calculating section 19. Noise variation Namh (cep), in which an estimated value Ha^(cep) of the multiplicative distortions that are obtained from voice that is actually uttered, additive noise Na(cep) that is obtained in a non-utterance period, and additive noise Nm(cep) of the initial noise HMM 17 are combined, is multiplied by a Jacobian matrix, wherein the result of the multiplication and initial combination HMM 16 are combined, and an adaptive HMM 26 is generated. Thereby, an adaptive HMM 26 that is matched to the observation value series RNah(cep) generated from actual utterance voice can be generated in advance.

Abstract: A multiplicative distortion Hm(cep) is subtracted from a voice HMM 5, a multiplicative distortion Ha(cep) of the uttered voice is subtracted from a noise HMM 6 formed by HMM, and the subtraction results Sm(cep) and {Nm(cep)?Ha (cep)} are combined with each other to thereby form a combined HMM 18 in the cepstrum domain. A cepstrum R^a(cep) obtained by subtracting the multiplicative distortion Ha (cep) from the cepstrum Ra (cep) of the uttered voice is compared with the distribution R^m(cep) of the combined HMM 18 in the cepstrum domain, and the combined HMM with the maximum likelihood is output as the voice recognition result.

Abstract: A trained vector creating part 15 creates a characteristic of an unvoiced sound in advance as a trained vector V. Meanwhile, a threshold value THD for distinguishing a voice from a background sound is created based on a predictive residual power ? of a sound which is created during a non-voice period. As a voice is actually uttered, an inner product computation part 18 calculates an inner product of a feature vector A of an input signal Sa and a trained vector V, and a first threshold value judging part 19 judges that it is a voice section when the inner product has a value which is equal to or larger than a predetermined value ? while a second threshold value judging part 21 judges that it is a voice section when the predictive residual power ? of the input signal Sa is larger than a threshold value THD.

Abstract: The digital signal conversion apparatus comprises: an over-sampling circuit that samples the input digital signal at high frequency; a polarity-inversion circuit that inverts the polarity of the sampled digital signal; interpolation circuit (A) and interpolation circuit (B) that perform interpolation of each respective digital signal; noise-shaping circuit (A) and noise-shaping circuit (B) that perform noise shaping on the interpolated signals; PWM conversion circuit (A) and PWM conversion circuit (B) that perform PWM conversion on the noise-shaped signals; and a switching circuit for driving a load based on the PWM signals from PWM conversion circuit (A) and PWM conversion circuit (B).

Abstract: A pulse width modulator for producing a PWM signal having reduced nonlinear distortion through interpolation processing with fewer computation steps is provided. The computational processing, W={1+&agr;(X2−X1)}{0.25 (X1+X2)}+0.5, is performed in accordance with successive sample values (X1, X2) in a PCM data train to determine a weighting factor (W). The computational processing, Xq=0.5·X0(W2−W)+X1(1−W2)+0.5·X2(W2+W), is performed using the sample values (X1, X2), a sample value (X0) previous to the sample value (X1), and the weighting factor (W) to thereby determine an interpolated sample value (Xq) having an amplitude close to that of an original analog signal (X(t)) generating the PCM data train. A point in time (tq) at which a reference signal (R(t)) takes on the interpolated sample value (Xq) is then determined to produce a PWM signal (Spwm) which is logically inverted at the point in time (tq).

Abstract: A delta-sigma modulator includes a quantization bit rate detection unit that detects a quantization bit rate of a digital audio signal, a volume setting value detection unit that detects a volume setting value of the digital audio signal; a filter that includes plural sets of filter coefficients having different shaping properties and allows the quantization noise to pass, and a filter coefficient switcher that switches the filter coefficients of the filter in accordance with the detection results of the quantization bit rate detection unit and the volume setting value detection unit. The delta-sigma modulator detects the quantization bit rate of an input signal and switches the filter coefficients to the low-order filter coefficient set when a source has a low quantization bit rate.

Abstract: The digital signal conversion apparatus comprises: an over-sampling circuit that samples the input digital signal at high frequency; a polarity-inversion circuit that inverts the polarity of the sampled digital signal; interpolation circuit (A) and interpolation circuit (B) that perform interpolation of each respective digital signal; noise-shaping circuit (A) and noise-shaping circuit (B) that perform noise shaping on the interpolated signals; PWM conversion circuit (A) and PWM conversion circuit (B) that perform PWM conversion on the noise-shaped signals; and a switching circuit for driving a load based on the PWM signals from PWM conversion circuit (A) and PWM conversion circuit (B).

Abstract: A voice recognition apparatus comprises a voice input device, a recognition processing device, a judging device and a setting device. The voice input device receives a voice input from a user. The recognition processing device performs a recognition processing to determine a plurality of word candidates corresponding to the voice input, through a matching processing with respective standby words in preset standby word groups. The judging device judges as whether or not the word candidates include a correct answer. The setting device determines a combination of most recognizable candidates in the word candidates and convertible word candidates thereof and sets same for the standby word groups to be used in a next recognition processing, in case where the judging device judges that the word candidate does not include the correct answer.

Abstract: An initial combination HMM 16 is generated from a voice HMM 10 having multiplicative distortions and an initial noise HMM of additive noise, and at the same time, a Jacobian matrix J is calculated by a Jacobian matrix calculating section 19. Noise variation Namh (cep), in which an estimated value Haˆ (cep) of the multiplicative distortions that are obtained from voice that is actually uttered, additive noise Na(cep) that is obtained in a non-utterance period, and additive noise Nm(cep) of the initial noise HMM 17 are combined, is multiplied by a Jacobian matrix, wherein the result of the multiplication and initial combination HMM 16 are combined, and an adaptive HMM 26 is generated. Thereby, an adaptive HMM 26 that is matched to the observation value series RNah(cep) generated from actual utterance voice can be generated in advance.

Abstract: A trained vector creating part 15 creates a characteristic of an unvoiced sound in advance as a trained vector V. Meanwhile, a threshold value THD for distinguishing a voice from a background sound is created based on a predictive residual power &egr; of a sound which is created during a non-voice period. As a voice is actually uttered, an inner product computation part 18 calculates an inner product of a feature vector A of an input signal Sa and a trained vector V, and a first threshold value judging part 19 judges that it is a voice section when the inner product has a value which is equal to or larger than a predetermined value &thgr; while a second threshold value judging part 21 judges that it is a voice section when the predictive residual power &egr; of the input signal Sa is larger than a threshold value THD.

Abstract: A multiplicative distortion Hm(cep) is subtracted from a voice HMM 5, a multiplicative distortion Ha(cep) of the uttered voice is subtracted from a noise HMM 6 formed by HMM, and the subtraction results Sm(cep) and {Nm(cep)−Ha (cep)} are combined with each other to thereby form a combined HMM 18 in the cepstrum domain. A cepstrum R&Lgr;a(cep) obtained by subtracting the multiplicative distortion Ha (cep) from the cepstrum Ra (cep) of the uttered voice is compared with the distribution R&Lgr;m(cep) of the combined HMM 18 in the cepstrum domain, and the combined HMM with the maximum likelihood is output as the voice recognition result.

Abstract: A dither generating apparatus which generates a sufficiently random auto dither even for a small-level signal and even if the buffer length is short. The LSB of quantized data is extracted and is stored in a buffer memory which serves as an M-bit shift register. An index buffer storing old 2.sup.M indexes is referred to with the M bits. A look-up table, which outputs a random value, is referred to in accordance with the value in the index buffer and outputs a dither. After the reference to the look-up table, current M-bit data is input to the index buffer so that the content of the index buffer is shifted piece by piece. As the values of the outputs of the index buffer and the buffer memory are both always variable, different values are always supplied to the look-up table, which therefore generates sufficiently random dithers.