Speech signal processing apparatus for detecting a speech signal from a noisy speech signal

A signal processing apparatus comprising:

• Fourier transformation means for Fourier-transforming an inputted speech signal and outputting a spectrum signal;
• cepstrum analysis means for performing a cepstrum analysis process for the spectrum signal outputted from said Fourier transformation means and outputting a cepstrum analysis result;
• pitch frequency detection means for detecting a pitch frequency in response to the cepstrum analysis result outputted from said cepstrum analysis means;
• pitch frequency judgment control means for storing predetermined pitch frequencies of speeches of plural groups of speech signals and sequentially outputting the stored pitch frequencies;
• pitch frequency judgment means for detecting which group among said plural groups of the speech signals the pitch frequency detected by said pitch frequency detection means corresponds to by comparing the pitch frequency detected by said pitch frequency detection means with the stored pitch frequencies outputted from said pitch frequency judgment control means and outputting a detection result;
• switch control means for outputting a switch control signal according to the detection result outputted from said pitch frequency judgment means;
• switch means for outputting spectrum signals discriminating them by respective groups of the speech signals according to the switch control signal outputted from said switch control means in response to the spectrum signal outputted from said Fourier transformation means;
• and
• inverse Fourier transformation means for inverse-Fourier-transforming the spectrum signals of respective groups of the speech signals outputted from said switch means and outputting the transformed speech signals of respective groups.

A signal processing apparatus comprising:

• band division means for performing a band division process including a Fourier transformation for an inputted speech signal and outputting spectrum signals of plural channels;
• cepstrum analysis means for performing a cepstrum analysis process for the spectrum signals of plural channels outputted from said band division means and outputting a cepstrum analysis result;
• pitch frequency detection means for detecting a pitch frequency in response to the cepstrum analysis result outputted from said cepstrum analysis means;
• pitch frequency judgment control means for storing predetermined pitch frequencies of speeches of plural groups of speech signals and sequentially outputting the stored pitch frequencies;
• pitch frequency judgment means for detecting which group among said plural groups of the speech signals the pitch frequency detected by said pitch frequency detection means corresponds to by comparing the pitch frequency detected by said pitch frequency detection means with the stored pitch frequencies outputted from said pitch frequency judgment control means and outputting a detection result;
• frequency control means for outputting a filter frequency control signal in response to the detection result outputted from said pitch frequency judgment means;
• comb type filter for filtering the spectrum signals of plural channels adjusting a filtering function under optimum conditions so as to pass only a speech signal thereof therethrough according to the filter frequency control signal in response to the spectrum signals of plural channels outputted from said band division means and outputting filtered spectrum signals of plural channels;
• and
• band combining means for combining the spectrum signals of plural channels filtered by said comb type filter, inverse-Fourier-transforming the combined spectrum signal, and outputting a transformed speech signal.

A signal processing apparatus comprising:

• band division means for performing a band division process including a Fourier transformation for an inputted noisy speech signal and outputting spectrum signals of plural channels;
• cepstrum analysis means for performing a cepstrum analysis process for the spectrum signals of plural channels outputted from said band division means and outputting a cepstrum analysis result;
• pitch frequency detection means for detecting a pitch frequency in response to the cepstrum analysis result outputted from said cepstrum analysis means;
• pitch frequency judgment control means for storing predetermined pitch frequencies of speeches of plural groups of speech signals and sequentially outputting the stored pitch frequencies;
• pitch frequency judgment means for detecting which group among said plural groups of the speech signals the pitch frequency detected by said pitch frequency detection means corresponds to by comparing the pitch frequency detected by said pitch frequency detection means with the stored pitch frequencies outputted from said pitch frequency judgment control means and outputting a detection result;
• frequency control means for outputting a filter frequency control signal in response to the detection result outputted from said pitch frequency judgment means;
• peak detection means for detecting a peak of a cepstrum in response to the cepstrum analysis result outputted from said cepstrum analysis means;
• speech judgment means for detecting a speech signal interval in response to the peak of the cepstrum detected by said peak detection means;
• noise prediction means for predicting a noise of the speech signal interval in response to the spectrum signals of plural channels outputted by said band division means according to the speech signal interval detected by said speech judgment means;
• cancellation means for cancelling the noise predicted by said noise prediction means in the spectrum signals of plural channels outputted from said band division means and outputting noise-cancelled spectrum signals of plural channels;
• comb type filter for filtering the spectrum signals of plural channels adjusting a filtering function under optimum conditions so as to pass only speech signals thereof therethrough according to the filter frequency control signal in response to the noise-cancelled spectrum signals of plural channels outputted from said cancellation means and outputting filtered spectrum signals of plural channels;
• and
• band combining means for combining the spectrum signals of plural channels filtered by said comb type filter, inverse-Fourier-transforming the combined spectrum signal, and outputting a transformed speech signal.

A signal processing apparatus comprising:

• band division means for performing a band division process including a Fourier transformation for an inputted noisy speech signal and outputting spectrum signals of plural channels;
• cepstrum analysis means for performing a cepstrum analysis process for the spectrum signals of plural channels outputted from said band division means and outputting a cepstrum analysis result;
• pitch frequency detection means for detecting a pitch frequency of a cepstrum in response to the cepstrum analysis result;
• formant analysis means for performing a formant analysis for the cepstrum analysis result outputted from said cepstrum analysis means and outputting a formant analysis result;
• speech detection means for detecting a feature of a speech in response to the pitch frequency detected by said pitch frequency detection means and the formant analysis result outputted from said formant analysis means;
• speech judgment control means for storing predetermined features of speeches of plural groups of speech signals and sequentially outputting the stored features of the speeches, said predetermined features denoting a pitch frequency and a formant analysis result for each of said plural groups of speech signals;
• speech judgment means for detecting which groups among plural groups of speech signals the feature detected by said speech detection means corresponds to by comparing the feature detected by said speech detection means with the features of the speeches outputted from said speech judgment control means, and outputting a detection result;
• frequency control means for outputting a filter frequency control signal in response to the detection result outputted from said speech judgment means;
• peak detection means for detecting a peak of a cepstrum in response to the cepstrum analysis result outputted from said cepstrum analysis means;
• speech judgment means for detecting a speech signal interval in response to the peak of the cepstrum detected by said peak detection means;
• noise prediction means for predicting a noise of the speech signal interval in response to the spectrum signals of plural channels outputted by said band division means according to the speech signal interval detected by said speech judgment means;
• cancellation means for cancelling the noise predicted by said noise prediction means in the spectrum signals of plural channels outputted from said band division means and outputting noise-cancelled spectrum signals of plural channels;
• comb type filter for filtering the spectrum signals of plural channels adjusting a filtering function under optimum conditions so as to pass only speech signals thereof therethrough according to the filter frequency control signal in response to the noise-cancelled spectrum signals of plural channels outputted from said cancellation means and outputting filtered spectrum signals of plural channels;
• and
• band combining means for combining the spectrum signals of plural channels filtered by said comb type filter, inverse-Fourier-transforming the combined spectrum signal, and outputting a transformed speech signal.

A signal processing apparatus comprising:

• band division means for performing a band division process including a Fourier transformation for an inputted noisy speech signal and outputting spectrum signals of plural channels;
• cepstrum analysis means for performing a cepstrum analysis process for the spectrum signals of plural channels outputted from said band division means and outputting a cepstrum analysis result;
• pitch frequency detection means for detecting a pitch frequency in response to the cepstrum analysis result outputted from said cepstrum analysis means;
• pitch frequency judgment control means for storing predetermined pitch frequencies of speeches of plural groups of speech signals and sequentially outputting the stored pitch frequencies;
• pitch frequency judgment means for detecting which group among said plural groups of the speech signals the pitch frequency detected by said pitch frequency detection means corresponds to by comparing the pitch frequency detected by said pitch frequency detection means with the stored pitch frequencies outputted from said pitch frequency judgment control means and outputting a detection result;
• switch control means for outputting a switch control signal according to the detection result outputted from said pitch frequency judgment means;
• peak detection means for detecting a peak of a cepstrum in response to the cepstrum analysis result outputted from said cepstrum analysis means;
• speech judgment means for detecting a speech signal interval in response to the peak of the cepstrum detected by said peak detection means;
• noise prediction means for predicting a noise of the speech signal interval in response to the spectrum signals of plural channels outputted by said band division means according to the speech signal interval detected by said speech judgment means;
• cancellation means for cancelling the noise predicted by said noise prediction means in the spectrum signals of plural channels and outputting noise-cancelled spectrum signals of plural channels;
• band combining means for combining the noise-cancelled spectrum signals of plural channels outputted by said cancellation means, inverse-Fourier-transforming the combined spectrum signal, and outputting a transformed speech signal;
• and
• switch means for outputting speech signals of respective groups discriminating them by respective groups of the speech signals according to the switch control signal outputted from said switch control means in response to the transformed speech signal outputted from said band combining means.

A signal processing apparatus comprising:

• band division means for performing a band division process including a Fourier transformation for an inputted noisy speech signal and outputting spectrum signals of plural channels;
• cepstrum analysis means for performing a cepstrum analysis process for the spectrum signals of plural channels outputted from said band division means and outputting a cepstrum analysis result;
• pitch frequency detection means for detecting a pitch frequency in response to the cepstrum analysis result outputted from said cepstrum analysis means;
• formant analysis means for performing a formant analysis process for the cepstrum analysis result outputted from said cepstrum analysis means and outputting a formant analysis result;
• speech detection means for detecting a feature of a speech of the inputted noisy speech signal in response to the pitch frequency detected by said pitch frequency detection means and the formant analysis result outputted from said formant analysis means;
• speech judgment control means for storing predetermined features of speeches of plural groups of speech signals and sequentially outputting the stored features of the speeches, said predetermined features denoting a pitch frequency and a formant analysis result for each of said plural groups of speech signals;
• speech judgment means for detecting which group among plural groups of speech signals the feature detected by said speech detection means corresponds to by comparing the feature detected by said speech detection means with the features of the speeches outputted from said speech judgment control means, and outputting a detection result;
• switch control means for outputting a switch control signal according to the detection result outputting from said speech frequency judgment means;
• peak detection means for detecting a peak of a cepstrum in response to the cepstrum analysis result outputted from said cepstrum analysis means;
• speech judgment means for detecting a speech signal interval in response to the peak of the cepstrum detected by said peak detection means;
• noise prediction means for predicting a noise of the speech signal interval in response to the spectrum signals of plural channels outputted by said band division means according to the speech signal interval detected by said speech judgment means;
• cancellation means for cancelling the noise predicted by said noise prediction means in the spectrum signals of plural channels outputted from said band division means and outputting noise-cancelled spectrum signals of plural channels;
• switch means for outputting speech signals of respective groups discriminating them by respective groups of the speech signals according to the switch control signal outputted from said switch control means in response to the noise-cancelled spectrum signals of plural channels outputted from said cancellation means;
• and
• band combining means for combining the spectrum signals of plural channels outputted by said switch means, inverse-Fourier-transforming the combined spectrum signal, and outputting a transformed speech signal.

A speech recognition apparatus comprising:

• first storage means for storing pitch frequencies of plural groups of speech signals which are previously obtained by performing a band division process including a Fourier transformation for speech signals of plural groups and performing a cepstrum analysis process for the transformed speech signal;
• band division means for performing a band division process including a Fourier transformation for an inputted noisy speech signal and outputting spectrum signals of plural channels;
• cepstrum analysis means for performing a cepstrum analysis process for the spectrum signals of plural channels outputted from said band division means and outputting a cepstrum analysis result;
• pitch frequency detection means for detecting a pitch frequency in response to the cepstrum analysis result outputted from said cepstrum analysis means;
• first comparison means for detecting which group the inputted noisy speech signal corresponds to by comparing the pitch frequency detected by said pitch frequency detection means with the pitch frequencies stored in said first storage means, and outputting a first comparison result;
• second storage means for storing predetermined features of speeches of plural groups of speech signals;
• speech cutting-out means for cutting out a speech from the inputted noisy speech signal;
• feature extraction means for extracting a feature of the speech cut out by said speech cutting-out means;
• second comparison means for recognizing a speech by comparing the feature of the speech extracted by said feature extraction means with the features of the speeches stored in the second storage means outputting a second comparison result;
• and
• comparison combining means for performing a predetermined process for the first comparison result outputted from said first comparison means and the second comparison result outputted from said second comparison means.