Abstract: A noise suppression system includes an a priori S/N ratio estimated value and expectation calculation unit that acquires an expectation of a priori S/N ratio, by correcting an estimated value of the a priori S/N ratio relating to a signal and a noise based on a priori S/N ratio model or based on a signal model and a noise model, the signal and the noise being estimated from an input signal in which the signal and the noise are mixed; a noise suppression coefficient calculation unit that calculates a noise suppression coefficient with use of the expectation of the a priori S/N ratio; and a noise suppression unit that suppresses the noise included in the input signal by multiplying the input signal by the noise suppression coefficient.

Abstract: A speech processing apparatus includes: an expectation value calculation unit configured to calculate, using an input signal spectrum and a speech model that models a feature quantity of speech, a spectrum expectation value which is an expectation value of a spectrum of an acoustic component included in the input signal spectrum; and an acoustic power estimation unit configured to estimate an acoustic power of the acoustic component of the input signal spectrum based on the input signal spectrum and the spectrum expectation value.

Abstract: To remove only noise components without removing desired signal components, a signal processing apparatus includes a noise decorrelator that removes noise signals having correlation between at least two input signals, in each of which a desired signal and a noise signal coexist, by receiving the at least two input signals from at least two channels, and a residual noise remover that removes residual noise included in an output signal of the noise decorrelator based on a phase difference between the output signal of the noise decorrelator and at least one input signal included in the at least two input signals.

Abstract: The purpose of the present invention is to provide a technology which is capable of appropriately evaluating a person's conduct with respect to another person. Provided is an information processing device, comprising a recognition unit 11, a detection unit 12, and an evaluation unit 13. The recognition unit 11 recognizes an evaluation subject's conduct. The detection unit 12 detects a trigger which is a state of a person other than the evaluation subject which triggers the evaluation subject's conduct. Using the detected trigger and the result of recognition by the recognition unit 13 relating to the evaluation subject's conduct, the evaluation unit 13 evaluates the evaluation subject's conduct.

Abstract: A speech processing apparatus includes: an expectation value calculation unit configured to calculate, using an input signal spectrum and a speech model that models a feature quantity of speech, a spectrum expectation value which is an expectation value of a spectrum of an acoustic component included in the input signal spectrum; and an acoustic power estimation unit configured to estimate an acoustic power of the acoustic component of the input signal spectrum based on the input signal spectrum and the spectrum expectation value.

Abstract: A noise suppression system includes an a priori S/N ratio estimated value and expectation calculation unit that acquires an expectation of a priori S/N ratio, by correcting an estimated value of the a priori S/N ratio relating to a signal and a noise based on a priori S/N ratio model or based on a signal model and a noise model, the signal and the noise being estimated from an input signal in which the signal and the noise are mixed; a noise suppression coefficient calculation unit that calculates a noise suppression coefficient with use of the expectation of the a priori S/N ratio; and a noise suppression unit that suppresses the noise included in the input signal by multiplying the input signal by the noise suppression coefficient.

Abstract: To remove only noise components without removing desired signal components, a signal processing apparatus includes a noise decorrelator that removes noise signals having correlation between at least two input signals, in each of which a desired signal and a noise signal coexist, by receiving the at least two input signals from at least two channels, and a residual noise remover that removes residual noise included in an output signal of the noise decorrelator based on a phase difference between the output signal of the noise decorrelator and at least one input signal included in the at least two input signals.

Abstract: A method for processing multichannel acoustic signals which processes input signals of a plurality of channels including the voices of a plurality of speaking persons. The method is characterized by detecting the voice section of each speaking person or each channel, detecting overlapped sections wherein the detected voice sections are common between channels, determining a channel to be subjected to crosstalk removal and the section thereof by use of at least voice sections not including the detected overlapped sections, and removing crosstalk in the sections of the channel to be subjected to the crosstalk removal.

Abstract: A method for processing multichannel acoustic signals, whereby input signals of a plurality of channels including the voices of a plurality of speaking persons are processed. The method is characterized by comprising: calculating the first feature quantity of the input signals of the multichannels for each channel; calculating similarity of the first feature quantity of each channel between the channels; selecting channels having high similarity; separating signals using the input signals of the selected channels; inputting the input signals of the channels having low similarity and the signals after the signal separation; and detecting a voice section of each speaking person or each channel.

Abstract: A speech recognition apparatus that performs frame synchronous beam search by using a language model score look-ahead value prevents the pruning of a correct answer hypothesis while suppressing an increase in the number of hypotheses. A language model score look-ahead value imparting device 108 is provided with a word dictionary 203 that defines a phoneme string of a word, a language model 202 that imparts a score of appearance easiness of a word, and a smoothing language model score look-ahead value calculation means 201. The smoothing language model score look-ahead value calculation means 201 obtains a language model score look-ahead value at each phoneme in the word from the phoneme string of the word defined by the word dictionary 203 and the language model score defined by the language model 202 so that the language model score look-ahead values are prevented from concentrating on the beginning of the word.

Abstract: The present invention enables the recognition process at high speed even when a lot of garbage is included in the grammar. The first voice recognition processing unit generates a recognition hypothesis graph which indicates a structure of hypothesis that is derived according to a first grammar together with a score associated with respective connections of a recognition unit by executing a voice recognition process based on the first grammar to a voice feature amount of input voice, and the second voice recognition processing unit outputs the recognition result from a total score of a hypothesis which is derived according to a second grammar after executing a voice recognition process according to the second grammar that is specified to accept a section other than keywords in input voice as the garbage section to a voice feature amount of input voice, and the second voice recognition processing unit acquires the structure and the score of the garbage section from the recognition hypothesis graph.

Abstract: A method for processing multichannel acoustic signals, whereby input signals of a plurality of channels including the voices of a plurality of speaking persons are processed. The method is characterized by comprising: calculating the first feature quantity of the input signals of the multichannels for each channel; calculating similarity of the first feature quantity of each channel between the channels; selecting channels having high similarity; separating signals using the input signals of the selected channels; inputting the input signals of the channels having low similarity and the signals after the signal separation; and detecting a voice section of each speaking person or each channel.

Abstract: A method for processing multichannel acoustic signals which processes input signals of a plurality of channels including the voices of a plurality of speaking persons. The method is characterized by detecting the voice section of each speaking person or each channel, detecting overlapped sections wherein the detected voice sections are common between channels, determining a channel to be subjected to crosstalk removal and the section thereof by use of at least voice sections not including the detected overlapped sections, and removing crosstalk in the sections of the channel to be subjected to the crosstalk removal.

Abstract: A speech recognition apparatus that performs frame synchronous beam search by using a language model score look-ahead value prevents the pruning of a correct answer hypothesis while suppressing an increase in the number of hypotheses. A language model score look-ahead value imparting device 108 is provided with a word dictionary 203 that defines a phoneme string of a word, a language model 202 that imparts a score of appearance easiness of a word, and a smoothing language model score look-ahead value calculation means 201. The smoothing language model score look-ahead value calculation means 201 obtains a language model score look-ahead value at each phoneme in the word from the phoneme string of the word defined by the word dictionary 203 and the language model score defined by the language model 202 so that the language model score look-ahead values are prevented from concentrating on the beginning of the word.

Abstract: The present invention enables the recognition process at high speed even when a lot of garbage is included in the grammar. The first voice recognition processing unit generates a recognition hypothesis graph which indicates a structure of hypothesis that is derived according to a first grammar together with a score associated with respective connections of a recognition unit by executing a voice recognition process based on the first grammar to a voice feature amount of input voice, and the second voice recognition processing unit outputs the recognition result from a total score of a hypothesis which is derived according to a second grammar after executing a voice recognition process according to the second grammar that is specified to accept a section other than keywords in input voice as the garbage section to a voice feature amount of input voice, and the second voice recognition processing unit acquires the structure and the score of the garbage section from the recognition hypothesis graph.