TH182015A - Deciding if it is an echo / echo for speech processing. - Google Patents

Abstract

According to the nature of this invention The methods for processing speech include judging. The non-echo / echo parameter that reflects the characteristics of the non-echo / echo in the current frame. Of speech signals, which are made up of frames Flatten / Flatten parameters Then a decision will be determined to include information of the non-echo / echo parameters in the frame. Address preceding the current frame of speech signals. The difference between the non-echo / echo parameter with Smoothed echo / smoothed parameters will be calculated. The method also includes procreation. Decide point that it is no echo / echo for judgment. The present frame is composed of words. That do not echo, or whether words are echoed by using calculated differences as a decision parameter

Claims (17)

------24/01/2562------(OCR) Page 1 of 4 pages. Claims 1. Speech Processing Method. The method comprises: determination of the resonance-nosing parameter for the first frame of a speech signal composed of multiple frames, where the resonance-nosing parameter is a composite parameter reflecting at least two characteristics of the resonant speech in the first frame; determination of the smoothed resonance-nosing parameter for the first frame, where this smoothed resonance-nosing parameter for the first frame is calculated from the smoothed resonance-nosing parameter for the second frame and the resonance-nosing parameter for the first frame, where the second frame is the preceding frame of the first frame; calculation of the difference between the resonance-nosing parameter for the first frame and the smoothed resonance-nosing parameter for the first frame; and determination of whether the first frame is a resonant speech signal by comparing the calculated difference to at least one threshold level. 2. Method of Claim 1, where at least two resonance characteristics of the voice are comprised of the periodic characteristic characteristic and the spectral tilt characteristic. 3. Method of Claim 2, where the composite parameter is calculated from the periodic characteristic parameter and the spectral tilt parameter. 4. Method of Claim 1, where the glass smoothing resonance parameter for the first frame is determined by weighting the resonance parameter for the first frame and the glass smoothing resonance parameter for the second frame, where if the glass smoothing resonance parameter for the second frame is greater than the resonance parameter for the first frame, the glass smoothing resonance parameter for the second frame is given less weight than if the glass smoothing resonance parameter for the second frame is not greater than the resonance parameter for the first frame. Page 2 of 4. 5. Method of Claim. 4. Where the weighting factor of the smoothing parameter for the second frame is 0.9 and the weighting factor of the smoothing parameter for the first frame is 0.1, when the smoothing parameter for the second frame is greater than the smoothing parameter for the first frame. Or where the weighting factor of the smoothed non-smoothing parameter for the second frame is 0.99 and the weighting factor of the non-smoothing parameter for the first frame is 0.01 when the smoothed non-smoothing parameter for the second frame is not greater than the non-smoothing parameter for the first frame. 6. Method of Claim 5 where the decision of whether the first frame is a non-smoothing speech signal or not by comparing the calculated difference to at least one threshold level consists of: a decision that the first frame is a non-smoothing speech signal when the calculated difference is greater than 0.1, or a decision that the first frame is not a non-smoothing speech signal when the calculated difference is less than 0.05. 7. Method of Claim 6 where the method also consists of a decision that the first frame has the same type of speech as the preceding frame of the first frame when the calculated difference is not less than 0.05. And no more than 0.1. 8. Method of claim 1, where the first and second frames are frames or subframes of a speech signal. 9. A speech processing unit consists of: a processor; and a non-temporal computer-readable storage medium that stores computer instructions which, when executed by the processor, cause the processor to perform the following operations: Page 3 of 4. Determine the resonance-nozing parameters for the first frame of a speech signal comprising multiple frames, where the resonance-nozing parameters are composite parameters reflecting at least two characteristics of the resonance-nozing speech in the first frame; Determine the smoothing-nosonance-nozing parameters for the first frame, where these smoothing-nosonance-nozing parameters for the first frame are calculated from the smoothing-nosonance-nozing parameters for the second frame, where the second frame is the frame preceding the first frame; Calculate the difference between the resonance-nozing parameters for the first frame and the smoothing-nosonance-nozing parameters for the first frame; And determine whether the first frame is a voiceless signal or not by comparing the calculated difference to at least one initial shift level.10. The suite of rules 9, where at least two voiceless characteristics are comprised of the periodic characteristic of the signal and the spectral tilt characteristic.11. The suite of rules 10, where the composite parameter is calculated from the periodic characteristic parameter and the spectral tilt parameter.12. Any one of the suites 9 to 11, where the glass smoothing voiceless parameter for the first frame is determined by weighting the voiceless parameter for the first frame and the glass smoothing voiceless parameter for the second frame, where if the glass smoothing voiceless parameter for the second frame is greater than the voiceless parameter for the first frame, the glass smoothing voiceless parameter for the second frame is given less weight than if the glass smoothing voiceless parameter for the second frame is not greater than the voiceless parameter for the first frame.13. The set of parameters of claim 12, where the weighting factor of the reverberation-free parameter for the second frame is 0.9 and the weighting factor of the reverberation-free parameter for the first frame is 0.1, when the reverberation-free parameter for the second frame is greater than the reverberation-free parameter for the first frame, or page 4 of 4. 13. The first frame is judged to be a voiceless speech signal when the calculated difference is greater than 0.1, or the first frame is judged not to be a voiceless speech signal when the calculated difference is less than 0.05. 14. The set of claims 9 through 11 where the first and second frames are frames or subframes of the speech signal. 17. An audio access device comprising an encoder and decoder containing an encoder or decoder, where the encoder or decoder is configured to execute one of the methods of Rights 1 through 8. 18. An audio access device of Rights 17 where the encoder or decoder is part of a digital signal processing (DSP) chip. 19. A set of devices of Rights 18 where the encoder and decoder are executed by software running on a processor or by task-specific hardware. 20. A computer-readable storage medium containing computer instructions which, when executed by a processor, cause the processor to execute one of the steps of Rights 1 through 8. 1. The speech processing method comprises: determining the devouring/reverberating parameters that reflect the characteristics of the devouring/reverberating speech in the current frame of the speech signal, which consists of multiple frames; determining the smoothed devouring/reverberating parameters to include information on the devouring/reverberating parameters in the frames preceding the current frame of the speech signal; calculating the difference between the devouring/reverberating parameters and the smoothed devouring/reverberating parameters; and determining whether the current frame contains devouring or reverberating speech using the calculated difference as a decision parameter. 2. The method described in Claim 1, where the non-reverberating/reverberating parameters are composite parameters that reflect at least two characteristics of non-reverberating/reverberating speech. 3. The method as specified in Claim 2, where the composite parameter is the product of the periodic characteristic parameter and the spectral tilt parameter. 4. Any of the methods specified in Claims 1 through 3, where the devoicing/reverberating parameter is a punvoicing parameter that reflects the characteristics of voiceless speech, and the smoothed devoicing/reverberating parameter is a smoothed punvoicing parameter (Punvoicing_sm). 5. The method specified in Claim 2 states that when the difference between the resonance-nosing parameter and the smoothed resonance-nosing parameter is greater than 0.1, the current frame of the speech signal is deemed resonant; and when the difference between the resonance-nosing parameter and the smoothed resonance-nosing parameter is less than 0.05, the current frame of the speech signal is deemed not resonant. 6. The method described in Claim 2 states that when the difference between the resonance-nosing parameter and the smoothed resonance-nosing parameter is between 0.05 and 0.1, the current frame of the speech signal is determined to have the same speech type as the previous frame. 7. The method according to one of claims 4 through 6, where the resonance-free parameter is calculated from the following resonance-free parameter: if (Punvoicing_sm > Punvoicing) { Punvoicing_sm (formula) 0.9 Punvoicing_sm + 0.1 Punvoicing } else { Punvoicing_sm (formula) 0.99 Punvoicing_sm + 0.01 Punvoicing } . 8. Any one of the claims 1 through 3 shall apply, whereby the devoicing/voicing parameter shall be the voice-reflective voice-reflective parameter (P voicing) that reflects the specific characteristics of the voiced speech, and whereby the smoothed devoicing/voicing parameter shall be the smoothed voice-reflective parameter (P voicing_sm). 9. The method specified in claim 8 states that when the difference between the resonant parameter and the smoothed resonant parameter is greater than 0.1, the current frame of the speech signal is determined to be resonant; and when the difference between the resonant parameter and the smoothed resonant parameter is less than 0.05, the current frame of the speech signal is determined to not be resonant. 10. The method specified in claim 8 or 9, where the smoothed resonance parameter is calculated from the resonance parameter as follows: if (Pvoicing_sm > Pvoicing) { Pvoicing_sm (formula) (7 / 8) Pvoicing_sm + (1 / 8) Pvoicing } else { Pvoicing_sm (formula) (255 / 256) Pvoicing_sm + (1 / 256) Pvoicing } . 11. Any one of the ten claims shall be implemented in a manner that includes sub-frames. 12. A speech processing unit comprising: a processor; and a computer-readable storage medium containing the programming for execution by the processor. This programming includes instructions to perform the following operations: determine the deverbinating/reverberating parameters that reflect the characteristics of the deverbinating/reverberating speech in the current frame of a speech signal composed of multiple frames; determine the smoothed deverbinating/reverberating parameters to include information on the deverbinating/reverberating parameters in frames preceding the current frame of the speech signal; calculate the difference between the deverbinating/reverberating parameters and the smoothed deverbinating/reverberating parameters; and determine whether the current frame contains deverbinating or reverberating speech using the calculated difference as a decision parameter. 13. The instrument set is described in Claim 12, where the resonance/resonance-causing parameters are composite parameters reflecting the product of the periodic characteristic parameter and the spectral tilt parameter. 14. The equipment specified in Claim 12 or 13 states that when the difference between the denouncing/reverberating parameter and the smoothed denouncing/reverberating parameter is greater than 0.1, the current frame of the speech signal is determined to be denouncing/reverberating; and when the difference between the denouncing/reverberating parameter and the smoothed denouncing/reverberating parameter is less than 0.05, the current frame of the speech signal is determined to be non-denouncing/reverberating. 15. Any set of instruments under claims 12 through 14 where the devouring/reverberating parameters are devouring parameters that reflect the characteristics of voiceless speech, and where the smoothed devouring/reverberating parameters are smoothed devouring parameters. 16. Any set of instruments under claims 12 through 14 where the devouring/reverberating parameters are reverberating parameters that reflect the characteristics of reverberating speech, and where the smoothed devouring/reverberating parameters are smoothed reverberating parameters. 17. A set of instruments pursuant to any one of claims 13 through 16, where the framework comprises sub-frames.