A kind of digital technology processing method of the appliance that is used to stutter
Technical field
The present invention relates to a kind of audio signal processing method, the digital technology processing method of definite a kind of appliance that is used to stutter of saying so being applicable to that stutter corrects or auxiliary treatment.
Background technology
Audition changes feedback (Altered auditory feedback is to call AAF in the following text) and is meant that the speaker hears self mode through the voice after changing, and it is a kind of verbal feedback method of utilizing the audition sense organ.AAF can have audition to postpone feedback (Delayed Auditory Feedback, to call DAF in the following text), frequency shift feedback (Frequency Altered Feedback, to call FAF in the following text) and shelter three kinds of modes of auditory feedback (masking auditory feedback is to call MAF in the following text).The stutter patient uses audition to change feedback system and speaks, and can improve speaker's smooth degree, alleviates the stutter symptom.DAF and FAF are the desirable systems of selection that instrument is corrected in stutter, and MAF is also useful to the patient that partly stutters, and especially is easy to produce the not people of the stutter behavior of sounding at those.It is to utilize human body mirror element nervous system (to see intuitively in essence that audition changes feedback, be to utilize the mechanism interim adaptation, neural that is used for imitating and helping neonate existence originally) obtain and developing effect at language, wait the communication disease except being used for improving stutter, research in recent years more relates to the research that waits the mirror-image system dysfunction as autism patient.
More existing at present instruments that adopt DAF to correct a stammer, they adopt memorizer (as magnetizing mediums), by the control of read-write sequence, realize the delay of voice signal.The shortcoming of this method is the restriction that is subjected to memory read/write cycle time delay, can not satisfy very brief, accurate requirement time delay that in recent years the DAF institute is proposed.Especially be easy to produce the not stutter behavior of sounding for those and the patient of lead-in dysphonia has no effect.
Summary of the invention
The objective of the invention is to deficiency, a kind of digital technology processing method of the appliance that is used to stutter is provided, being applicable to that stutter corrects or auxiliary treatment at existing stutter appliance and processing method thereof.The present invention adopts digital technology, voice signal is carried out audition postpone feedback, frequency shift feedback and shelter three kinds of feedback processing of auditory feedback.Greatly improve speaker's the smooth degree of pronunciation, alleviate the stutter symptom, especially the patient's effect for stutter behavior that is easy to produce sounding not and lead-in dysphonia is remarkable.
For achieving the above object, the technical solution used in the present invention is as follows:
A kind of digital technology processing method of the appliance that is used to stutter is characterized in that: steps of the method are:
A, signal input: pick off becomes level signal with analog signal conversion, and A/D converter A/D converts level signal to digital signal, and then adopts Fourier transform, and time/frequency domain converter is transformed into frequency-region signal with digital signal from time-domain signal;
B, signal identification and optimization: the frequency-region signal after conversion is implemented the algorithm noise reduction, calculate the letter/specific factor of making an uproar of each wave band of frequency-region signal, decision should give the attenuation of the mixed noise cancellation signal of each frequency range and the amount of gain of voice signal, the noise signal of each frequency range that decays, the voice signal of each frequency range that gains.
C, signal reconstruction: frequency-region signal delayed wave filter simultaneously carries out audition delay feedback processing with frequency-region signal, through frequency modifier frequency-region signal is carried out the frequency shift feedback processing;
D, shelter optimization: the frequency-region signal after audition postpones feedback and frequency shift feedback processing carries out inverse-Fourier transform, frequently/the time domain transducer converts frequency-region signal to time-domain signal, time-domain signal is discerned through recognizer, the masking signal that the control mixer will be sheltered the maker generation mixes output with this time-domain signal, finishes and shelters the auditory feedback processing;
E, signal output: the digital signal after audition postpones feedback processing, frequency shift feedback processing and shelters the auditory feedback processing converts level signal to through D/A D/A, is exported by pick off.
It is by delay filter frequency-region signal to be postponed to handle that described audition postpones feedback processing.
Described frequency shift feedback processing is by frequency modifier frequency-region signal to be carried out frequency shift to handle: the frequency domain components of each wave band of described frequency-region signal is according to frequency height sequence arrangement, and described frequency modifier rearranges described frequency domain components by specified putting in order.
Described masking signal is white noise or narrow-band noise, and described white noise is generated by the random function algorithm, and described narrow-band noise is the pass-band noise of white noise through producing behind the band filter.
Described pick off is mike or speaker.
The invention has the advantages that:
1, the present invention adopts audition to postpone feedback, frequency shift feedback and shelters three kinds of feedbacks of auditory feedback acoustical signal is handled, can effectively improve the smooth degree of stutter patient speech, treatment stutter more remarkable effect, stable, applied widely, more existing stutter therapeutic instrument effect of only stuttering by the inhibit signal treatment is more obvious, especially the people's effect for people who is easy to produce the not stutter behavior of sounding and lead-in dysphonia is remarkable.
2, the present invention adopts the mode that digital technology is handled, and required number of elements is few, volume is little, low in energy consumption, is convenient to adopt the present invention to develop minitype portable equipment, to satisfy user to outward appearance and concealed demand.
3, the present invention is by being digital signal with analog signal conversion, be converted to frequency-region signal again, frequency-region signal is postponed and frequency shift, the advanced algorithm of Applied Digital signal processing easily in this process is realized high-quality, real-time signal reconstruction again.
4, the delay auditory feedback that studies show that minimum 4ms just begins that user is improved the language fluency degree and works, and time delay is short more, more little to the speed influence that user is spoken, therefore, the quick reconfiguration to signal that realizes by the present invention, accurate control are significant to the treatment stutter and the auxiliary treatment that is used to stutter.
Accompanying drawing and drawing explanation
Fig. 1 is an overall structure schematic flow sheet of the present invention
The schematic flow sheet that Fig. 2 optimizes for the identification of signal of the present invention
Fig. 3 is the schematic flow sheet of signal reconstruction of the present invention
Fig. 4 postpones the feedback processing principle schematic for signal of the present invention carries out audition
Fig. 5 carries out frequency shift feedback processing principle schematic for signal of the present invention
Fig. 6 generates the masking signal principle schematic for masking signal maker of the present invention
Fig. 7 shelters auditory feedback handling principle sketch map for signal of the present invention
The specific embodiment
As shown in Figure 1: signal flow of the present invention is divided into the following stage:
Input signal: pick off such as mike become level signal with analog signal conversion, A/D converter (A/D) converts level signal to digital signal and also promptly mixes noisy speech signal, and then the employing Fourier transform, time/frequency domain converter is transformed into frequency-region signal with digital signal from time-domain signal;
Signal identification and optimization: the frequency-region signal after conversion is implemented the algorithm noise reduction, calculate the letter/specific factor of making an uproar of each wave band of frequency-region signal, decision should give the attenuation of the mixed noise cancellation signal of each frequency range and the amount of gain of voice signal, the noise signal of each frequency range that decays, the voice signal of each frequency range that gains.
Signal reconstruction: frequency-region signal delayed wave filter simultaneously carries out audition delay feedback processing with frequency-region signal, through frequency modifier frequency-region signal is carried out the frequency shift feedback processing;
Shelter optimization: the frequency-region signal after audition postpones feedback and frequency shift feedback processing carries out inverse-Fourier transform, frequently/the time domain transducer converts frequency-region signal to time-domain signal, time-domain signal is discerned through recognizer, the masking signal that the control mixer will be sheltered the maker generation mixes output with this time-domain signal, finishes and shelters the auditory feedback processing;
Signal output: the digital signal after audition postpones feedback processing, frequency shift feedback processing and shelters the auditory feedback processing converts level signal to through D/A D/A, is exported by pick off such as speaker.
As shown in Figure 2: the identification of signal of the present invention and optimization step are:
1, the time/and frequency domain conversion: adopt fast Fourier transform (FFT), will mix the voice of making an uproar and be transformed into frequency-region signal, obtain mixing the frequency component (N by FFT count decision) of the voice N wave band of making an uproar from time-domain signal.
2, specific factor is believed/is made an uproar in calculating: press certain hour at interval, calculate each wave band instantaneous energy value; By certain hour interval (time is the several times at aforementioned interval), in several instantaneous energy values, calculate its maximum and minima, and record; According to maximum and minima computation of characteristic values as believing/make an uproar specific factor, it is eigenvalue=f (maximum, minima) (computational algorithm of eigenvalue has multiple, as to ask its meansigma methods be exactly wherein a kind of, adopt which kind of algorithm can by at noise type, by a learning algorithm or rule of thumb determine.)
3, be attenuation relation, amount of gain relation according to the letter/specific factor of determining by learning algorithm or experience in advance of making an uproar, calculate in this interval to the attenuation and the amount of gain of this band signal, the letter of mixed noise cancellation signal/make an uproar than high more believes/makes an uproar that the numerical value of specific factor is big more, and attenuation is more little.(variation of the signal of consideration is normally successive, gradual change, in the practical application, can be the attenuation that calculates in the next interval.)
4, each band signal in this interval is decayed respectively and using gain (size of gain is by amplifying the strategy decision, as adopting EDRC wide dynamic range Compression Strategies etc.).Thereby the noise in the signal is suppressed, and the signal envelope of voice is restored from noise, has improved the letter/ratio of making an uproar.
5, carrying out other handles as signal being carried out audition delay feedback, frequency shift feedback processing and sheltering auditory feedback and handle.
6, frequently/and the time domain conversion: adopt fast fourier transform (IFFT), the signal after optimizing is transformed into time-domain signal from frequency-region signal
As shown in Figure 4, the operation principle that signal of the present invention carries out delay filter in the audition delay feedback processing is: delay filter is based on an annular inputoutput buffer, the frequency-region signal of input always is placed on the position of input pointed, and output signal is always obtained from the position that output pointer points to.Suppose that output pointer is from original position, and the input pointer is before output pointer, N position (the data representative intervals of specified time delay/one of N=time input and output), interval, the input and output pointer travels forward synchronously, like this, and the signal of each input, will after specified blanking time, be output, reach the purpose of signal delay, adjust the interval of length, input pointer and the output pointer of whole buffer, can set time delay arbitrarily.
As shown in Figure 5: each frequency component has been moved down a wave band by integral body, and the frequency component 1 of wave band 1 is moved to wave band 2, and former wave band 1 fills out 0; The frequency component 2 of wave band 2 is moved to wave band 3; By that analogy, be moved to wave band n until the frequency component n-1 of wave band n-1, and the frequency component of former wave band n abandons.The new arrangement that obtains is like this compared with the original signal frequency, and its frequency rising is known clearly (supposing from wave band 1 to wave band n it is to arrange by the frequency ascending order), and the frequency of rising is the bandwidth of a wave band.Suppose that a wave band bandwidth is 500Hz, the original signal frequency is the 1000Hz signal, then obtains the signal that signal is frequency 1500Hz.
As shown in Figure 6: the generation of masking signal:
Masking signal is white noise or narrow-band noise, and white noise available random function generates, and promptly average is 0, and variance is 1 gaussian random noise.Consider in the practical application that the restriction of the operational capability of digital signal processor adopts the white noise signal that records in advance as the white noise sound source among the present invention.Narrow-band noise is the pass-band noise that white noise produces through band filter.Among the present invention with white noise through after the frequency analysis, only get the frequency component of some or several adjacent wave bands, and the frequency component of all the other wave bands put 0, just can obtain narrow-band noise.
As shown in Figure 7: sheltering the auditory feedback treatment step is:
Recognizer checks whether the signal after the reconstruct is the speech signal, to control mixer only when the input of speech signal is arranged, masking signal is mixed output with the speech signal, otherwise directly export the signal (promptly when the speaker mourns in silence, not adding masking signal) after the reconstruct.The recognition methods of recognizer remove to adopt with aforementioned signal optimizing in the speech recognition algorithm mentioned mutually the method one of myopia adopt to calculate and believe/make an uproar specific factor and judge that whether it is above preset threshold.
Except control mixer output mixed signal by recognizer, can also directly control mixer output mixed signal or masking signal by control signal.This mainly is in order to allow user can pass through an external interface (as a button or remote controller), and a masking signal is sent out by system in control this method, and this masking signal can be used as and helps user to solve the stimulus signal of difficult pronunciation.This stimulus signal can also adopt the voice signal of recording except with the masking signal, as vowel " a ", to reach better effect.
The invention is not restricted to the foregoing description; in the design scope that claim of the present invention limited; the one of ordinary skilled in the art also can do some conspicuous changes to the foregoing description, but these changes all should fall within the protection domain of claim of the present invention.