CN100514447C - Signal proces and its modular set - Google Patents
Signal proces and its modular set Download PDFInfo
- Publication number
- CN100514447C CN100514447C CNB2004100489673A CN200410048967A CN100514447C CN 100514447 C CN100514447 C CN 100514447C CN B2004100489673 A CNB2004100489673 A CN B2004100489673A CN 200410048967 A CN200410048967 A CN 200410048967A CN 100514447 C CN100514447 C CN 100514447C
- Authority
- CN
- China
- Prior art keywords
- frequency domain
- domain data
- frequency
- signal processing
- phase
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Fee Related
Links
Images
Landscapes
- Circuit For Audible Band Transducer (AREA)
Abstract
A method for processing signal converts original signal to frequency domain data containing multiple frequency, phase and intensity and carries out frequency domain process at the same frequency band range. It includes details as arranging frequency domain on variable axis, adjusting all frequency domain data to assigned phase and carrying out displacement on variable axis to obtain new sequence for combination output according to variable generated on axis at time when signal phase is varied under this frequency band.
Description
[technical field]
The present invention relates to a kind of signal processing method, particularly relate to a kind of signal processing method that can suitably preserve and transmit the signal phase characteristic.
[background technology]
In the past for the processing mode of digital signals such as sound, image, often utilized Hilbert conversion or fast fourier transform (FFT), data-switching is got the intensity and the phase propetry of frequency domain, transmit after treatment.Maybe can only transmit under the situation of particular phases but can't transmit phase place, often change the original signal feature in the processing procedure at some.
With of the processing of artificial electronic ear for voice signal, be subject to the excitation mode of human nerve and ganglion cell (ganglion cell), technology all can only be transmitted the strength characteristics (" Wilson; B.S.et al.Nature 352; 236-238 ", " Smith et al; Nature 416 (6876): 87-90; March 7; 2002 ") of frequency domain at present, and extract important phonetic feature-for example transmit fundamental frequency and resonance peak or look for bigger frequency-region signal, just cast out reluctant phase propetry and people and think unessential information.Yet, distinguish that in central nervous system the phase propetry under each frequency is important identification factor really, should give complete preservation transmission in the projects such as the language meaning contains, music string rule, source of sound, audio direction.
Below enumerate several open disposal routes of using:
(1) only transmits the friendship frequency domain data of (cross) certain phase place more.
With reference to No. the 6480820th, United States Patent (USP), its processing mode be with voice signal change behind a plurality of frequency domain datas, casting out does not have axially to hand over the frequency domain data of (Axis crossing) more.Yet in fact, not having axially to hand over frequency domain data more at low frequency (below about 2000Hz) is the major part that accounts for whole voice signal, and casting out does not have axially to hand over frequency-region signal more, will omit most acoustic information.In addition, may not be certain under certain analysis rate, to have smooth-going axial friendship more,, produce this with interpolation method and axially hand over frequency domain data more, will make the sound generating distortion if by between great two frequency domain datas of phase differential at (especially high frequency) in the practical application.
(2) the capable ripple of the interior basilar memebrane (basilar membrane) of cochlea on the simulation physiology.
Consult U.S. Patent Publication No. 20030171786, voice signal obtains many complex datas via FFT, includes the frequency intensity and the phase propetry of signal.Its processing mode is the carrying out and the delay phenomenon of simulating capable ripple as far as possible, makes all output signals stimulate cochlea at more correct time point.Therefore but its handled complex data is cast out imaginary part at the very start, has just cast out phase propetry and part strength characteristics, has changed the original signal feature, and is final and can't reduce correct voice signal.
(3) stimulate each electrode simultaneously and rapidly.
Consult United States Patent (USP) No. 6594525, its special relationship that exists between each mid-band frequency at the stimulus modality of each electrode of electronic ear and voice signal is especially handled, cast out phase propetry but processing procedure is still, changed original sound signal section feature.
Same, consult United States Patent (USP) No. 6647297, be applied in the signal processing mode of artificial electronic eyes aspect in the past, in signal of video signal conversion transmittance process, also be to transmit the phase propetry of image on spatial frequency, be directly to be considered as equiphase to handle, therefore can only keep the strength characteristics that lumen (brightness) or color etc. stimulate.
Yet no matter be voice signal or signal of video signal, phase propetry is the greatest factor (Oppenheim, A.V. " Theimportance of phase in signals ", Proc.IEEE.69,529-541,1981) that original signal keeps sharpness really.If can suitably preserve and transmit the phase propetry of each frequency, then can improve the identification degree of sound or image in a large number.
In addition, with regard to OFDM (OFDM) modulating system, or other need the system of the information of transmitting with leggy, be necessary to develop the signal processing mode that the script multi-phase signals is integrated into specific phase angle equally, to transmit greater amount information.
[summary of the invention]
Therefore, the objective of the invention is a kind of signal processing method that can suitably preserve and transmit the phase propetry of each frequency is being provided.
So signal processing method of the present invention is to handle at a plurality of frequency domain datas in the identical frequency band scope, each frequency domain data includes intensity and phase propetry.This method comprises the following step:
A, frequency domain data is arranged on the variable axle, this variable axle can be time shaft or spatial axes, view axis;
B, with the phase propetry of frequency domain data, the frequency characteristic of its place frequency band, or aforementioned two specific characters are parameter simultaneously, be defined as phase parameter, frequency parameter, be used for adjusting the phase place of this frequency domain data at least one appointment phase angle, and change its position on this variable axle, its position change amount is equivalent under its frequency characteristic condition, during this frequency domain data phase change, and corresponding variable quantity in the required generation of this variable axle; And
C, will adjust to the frequency domain data addition of vectors of same position on the variable axle, or with other mathematic(al) mode computings combinations, for adjusting output.
[description of drawings]
About aforementioned and other technology contents, characteristics and effect of the present invention, in the following detailed description that cooperates with reference to a graphic preferred embodiment, can clearly understand.
Fig. 1-the 1st uses the synoptic diagram of the artificial electronic ear system of signal processing method of the present invention.
Fig. 1-2 is figure time delay of corresponding 22 electrode positions of the capable ripple of human body basilar membrane.
Fig. 2 is the present invention's first preferred embodiment flow chart of steps.
Fig. 3-the 1st, this first preferred embodiment receives voice signal figure.
Fig. 3-the 2nd utilizes the present invention to handle the voice signal figure of back reduction.
Fig. 4-the 1st, frequency-intensity map that voice signal obtains through fast fourier transform.
Fig. 4-the 2nd, voice signal obtains frequency-phase diagram through fast fourier transform.
Fig. 5-1 is the distribution plan of one first frequency band on time shaft.
Fig. 5-2 is one second frequency band distribution plans on time shaft.
Fig. 6-the 1st, the distribution plan after this embodiment first frequency band tagmeme adjustment on time shaft.
Fig. 6-the 2nd, the distribution plan after this embodiment second frequency band tagmeme adjustment on time shaft.
Fig. 7 is after variation example 1 method of this first band utilization tagmeme set-up procedure is adjusted, the distribution plan on time shaft.
Fig. 8-the 1st, variation example 3 methods of this first band utilization tagmeme set-up procedure, vector decomposes the composition synoptic diagram.
After Fig. 8-the 2nd, Fig. 8-1 adjust and finish, the distribution plan on time shaft.
Fig. 9 is the synoptic diagram of the frequency domain data of a non-perpendicular time shaft.
Figure 10-the 1st utilizes variation example 3 methods of tagmeme set-up procedure to handle Fig. 9 frequency domain data, distribution plan after the tagmeme adjustment.
Figure 10-the 2nd, the frequency domain data of Figure 10-1 add the distribution plan after the time change amount.
There is the synoptic diagram of complex frequency domain data in Figure 11-the 1st on a time point.
Figure 11-the 2nd, the frequency domain data of Figure 11-1 is through variation example 3 methods of tagmeme set-up procedure, the distribution plan after the processing.
Figure 12-the 1st, this first frequency band carry out the distribution plan after the vector stack step.
Figure 12-the 2nd, this second frequency band carry out the distribution plan after the vector stack step.
Figure 13 is the distribution plan after Figure 12-1 adjusts the intensity step.
Figure 14 is the resulting passage integral data of one the 22 a passage midband addition of vectors synoptic diagram.
Figure 15 be Figure 14 divided by the frequency band number in the passage, be the average frequency domain data of all frequency bands in this passage.
Figure 16-the 1st, Figure 15 and frequency characteristic are F
1Frequency band, carry out the frequency domain data distribution plan after tagmeme adjustment and the vector stack step.
Figure 16-the 2nd, Figure 15 and frequency characteristic are F
2Frequency band, carry out the frequency domain data distribution plan after tagmeme adjustment and the vector stack step.
Figure 17 is a signal processing apparatus block schematic diagram of the present invention.
Figure 18 is signal processing method second preferred embodiment of the present invention, uses the synoptic diagram of artificial electronic eyes of the present invention system.
Figure 19 is signal processing method the 3rd preferred embodiment of the present invention, uses the spectrogram of orthogonal FDM modulation system of the present invention.
Figure 20-1 ~ 4th, the handled one first group of frequency domain data of the 3rd preferred embodiment distribution plan on time shaft.
Figure 21-1 ~ 4th, the handled one second group of frequency domain data of the 3rd preferred embodiment distribution plan on time shaft.
Figure 22-1 ~ 4th, Figure 20-1 ~ 4 is to specify phase angle 0 and π, the frequency domain data distribution plan after tagmeme adjustment and vector stack step process.
Figure 23-1 ~ 4th, Figure 21-1 ~ 4 is to specify phase angle 0.5 π and 1.5 π, the frequency domain data distribution plan after tagmeme adjustment and vector stack step process.
[embodiment]
Below in conjunction with drawings and Examples signal processing method of the present invention and device thereof are elaborated
As Figure 1-1, first preferred embodiment of signal processing method of the present invention is the signal processing that is applied in artificial electronic ear system 2.Artificial electronic ear system 2 comprises mainly that in simple terms one is used for capturing microphone 21, the speech processor 22 in order to processing audio signal 20 of voice signal 20, and the tandem electrode 23 of an implantation cochlea.Tandem electrode 23 directly stimulates cochlea after voice signal 20 is converted to electric wave, and is passed to brain through auditory nerve, complete deafness patient is set up or recovery hearing.Former electrodes 23 implant sites do not exceed with cochlea, implantablely listen the district in auditory nerve, brain stem or brain yet.
At the described tandem electrode 23 of present embodiment is to illustrate with 22 channel electrodes, and 22 electrodes are implanted the cochlea ad-hoc location, can correspondingly receive specific frequency range.For instance, plant the voice signal that can receive the highest frequency scope in the electrode of cochlea lowermost end (base), be defined as first passage; Plant the voice signal that then receives the low-limit frequency scope in the electrode of cochlea tip (apex), be defined as the 22 passage.Owing to calculate outward from the tip in the human cochlea, its distance with represent receive frequency to be logarithm to change (Greenwood ' s formula), so divide the 22 to first passage frequency bound with logarithm.The lower bound of supposing the 22 channel frequence scope is 125Hz, the upper bound of first passage frequency range is 8000Hz, and then the 22 logarithm frequency center to first passage is 147Hz, 196Hz, 253Hz, 320Hz, 397Hz, 488Hz, 594Hz, 718Hz, 863Hz, 1031Hz, 1229Hz, 1459Hz, 1727Hz, 2040Hz, 2406Hz, 2834Hz, 3332Hz, 3914Hz, 4594Hz, 5387Hz, 6313Hz, 7394Hz in regular turn.
The voice signal that human cochlea receives by importing place (corresponding to the electrode of first passage) into, reaches cochlea tip (corresponding to the electrode of the 22 passage) to go waveshape, therebetween the relation object of each electrode and time delay Fig. 1-2 seemingly.The signal processing mode of therefore artificial electronic ear is good can simulate row ripple transmission kenel, hereinafter will have the present invention to simulate the example explanation of capable ripple.
As shown in Figure 2, first preferred embodiment of signal processing method of the present invention comprises following steps:
Step 11-acquisition signal.
Cooperate and to consult Fig. 3-1, this step is by the collected a string simulated sound of 22 pairs of microphones of speech processor 21, obtains a digital audio signal 20 with the sampling rate that is higher than two times of sound frequencies that desire handles.Because the important audio frequency range of human daily life is between 125Hz ~ 8000Hz, so present embodiment adopts the 22050Hz sampling rate and explains, but frequency range of handling and sampling rate be not all as limit.
Step 12-frequency conversion process.
Present embodiment utilizes the fast fourier transform (FFT) of 256 number of samples that the voice signal among Fig. 3-1 20 is carried out frequency inverted, but mode processing signals such as also available Bart Wo Shifa (Butterworth) filtering, Hilbert conversion.
Present embodiment is adopted each sample of signal point and is made a FFT, therefore per second can obtain 22050 sections frequency spectrums, 256 frequency domain datas of output in just per 1/22050 second, every frequency domain data is R (f with the complex representation, t)+I (f, t) i can be considered a vector, includes the intensity (magnitude) and the phase place of this section voice signal 20 in this frequency) information.Wherein, parameter f is the representative frequency of this frequency domain data, and t is the position of this frequency domain data on the variable axle, is exactly time shaft at present embodiment variable axle
Cooperate and consult Fig. 4-1, Fig. 4-2.On behalf of this voice signal 20,256 numerical value that are arranged on the frequency axis among Fig. 4-1 carry out the intensity distributions of a resultant voice signal 20 of FFT on frequency axis, is defined as the 1st frequency F by the low frequency tremendously high frequency
1~ the 256 frequency F
256, each is to there being sectional frequency range, and subsequent content is with the F that mentions
1~ F
256Represent frequency to be meant the linear center frequency of each band frequency scope, entire scope is about 125Hz ~ 8000Hz.Fig. 4-1 is the frequency-intensity distribution of this voice signal 20, and the intensity on every Frequency point is equivalent to vector length √ (R
2+ I
2); Fig. 4-2 is corresponding frequency-phase change figure, and the phase value on every Frequency point equals tan
-1(I/R).
Shown in Fig. 5-1, Fig. 5-2, for convenience of explanation, below FFT is obtained the frequency domain data of characteristic frequency, be arranged on the time shaft according to sequential.At present embodiment, the unit spacing of time shaft | T
1|=| T
2-T
1|=| T
3-T
2| ...=1/22050 second.In addition the vertical axes of two vertical time axles, go out the paper azimuth axis and then represent imaginary number (I) line, real number (R) line respectively.And producing under the situation of frequency domain data with Bart Wo Shifa filtering, Hilbert conversion, this two axis then can be considered the footpath coordinate system of expression frequency domain data phase place and intensity.
For instance, Fig. 5-1 shows that continuous 6 sections FFT obtain F respectively
1 Frequency domain data 31 ~ 36, arrange according to the time and to obtain one first group of frequency band 30, represent this frequency sound variation in time.T
1F among the vector length presentation graphs 4-1 of frequency domain data 31 on the time point
1Intensity; The phasing degree of frequency domain data 31 is F among the presentation graphs 4-2 then
1Phase place.T
2~ T
6 Frequency domain data 32 ~ 36 on the time point is represented the resultant F of the second ~ six section FFT (figure does not show) respectively
1Intensity and phase place.
Fig. 5-the 2nd shows the frequency domain data 41 ~ 46 of the 2nd frequency that continuous 6 sections FFT obtain respectively, arranges according to the time and obtains one second group of frequency band 40.Be that explanation and comparison are convenient, with second group of frequency band 40 and identical the explaining of first group of frequency band, 30 distributions, but virtual condition usually can be not identical at this.T among the figure
1F among frequency domain data 41 length performance Fig. 4-2 on the time point
2Strength information; F among the frequency domain data 41 phasing degree representative graph 4-2
2Phase information.T
2~ T
6 Frequency domain data 42 ~ 46 on the time point shows the resultant F of the second ~ six section FFT (figure does not show) respectively
2Intensity and phase information.
In like manner can try to achieve the 3rd ~ the 256 group of frequency band, each frequency band is respectively represented certain frequency range, and the frequency band in the upper and lower boundary of the frequency of aforementioned 22 passages scope can find its respective channel, so each passage can receive at least one frequency band.
Step 13-tagmeme adjustment.
Consult Fig. 5-1, Fig. 5-2 and Fig. 6-1, Fig. 6-2 simultaneously, in the present embodiment for the phase information in the voice signal 20 is preserved, and the key character in unlikely deletion in signal processing, the loss voice signal 20, cause error, therefore step 13 according to sound under this frequency, relative required time when phase place changes, all frequency domain datas on the time shaft are adjusted into single phase place, and at the time shaft top offset, obtain new sequential and arrange, realize the target of fully compression and complete transmission.
What deserves to be mentioned is that the following treatment step of present embodiment can only be adjusted at the frequency domain data of audible frequency range of human body and intensity, human inaudible high frequency or low decibel signal then need not to adjust processing.Certainly, the filtration of also can going ahead of the rest in step 12 is only handled desire or the frequency domain data of audible frequency range of human body and intensity is arranged on the time shaft.
Set and specify phase angle Θ
0, specify phase angle Θ
0Can be arbitrarily angled, at present embodiment to specify single Θ
0Do explanation for=0 °.
Calculate the adjustment amount of this frequency domain data at the time shaft top offset.With T
1 Frequency domain data 31 on the time point can see through following [formula one] for instance, obtains the minimum adjustment amount Δ T that need produce on time shaft when phase place is adjusted.
Δ T=(Δ Θ/2 π) * T [formula one]
With regard to first group of frequency band 30, period T is 1/F
1, phase difference value Δ Θ is this frequency domain data and appointment phase angle Θ
0Between difference, the phase difference value Δ Θ=pi/2 of frequency domain data 31.Convenient for marginal data, suppose T=12|T
1|, then at this Δ T=3|T
1|, just output time point is at T
4
What deserves to be mentioned is that if specified non-single phase place in the step 131 is assumed to be 0 ° and π, then phase difference value Δ Θ is the former phase place of frequency domain data and specifies between phase angle the minimal difference of changeing (or complete in changeing clockwise) counterclockwise.
Change.At present embodiment is to carry out the conversion of phase place/time, frequency domain data 31 ~ 36 phase places all on Fig. 5-1 time shaft is adjusted to specified 0 ° in phase angle, and move to the output time point that step 132 is obtained, and obtains frequency band 30 ' after the adjustment shown in Fig. 6-1.
That is to say,, can make frequency domain data 31,32,33 on Fig. 5-1 be adjusted to 0 ° in appointment phase angle shown in Fig. 6-1, simultaneously respectively by time point T by [formula one] phase place/time conversion formula
1, T
2, T
3Be converted to T
4, T
8, T
5On, become frequency domain data 31 ', 32 ', 33 '.As among Fig. 5-1 originally phase place be 0 ° frequency domain data 34,36, or intensity is that zero frequency domain data 35 (not shown) then need not be changed.
According to identical thin step 131 ~ 133, the frequency domain data 41 ~ 46 among Fig. 5-2 on all time points can be adjusted to the single phase place shown in Fig. 6-2 by [formula one] phase place/time conversion formula.At present embodiment is that marginal data is convenient, is hypothesis F
2=2F
1, so period T=6|T
1|, though make second group of frequency band 40 identical, produce different adjustment amount Δ T with the frequency domain data intensity and the phase propetry of first group of frequency band 30, just output time point is different, and the output time point of frequency domain data 41 ~ 43 is respectively at T
2With T
3Between, T
5Upward, T
4On, obtain frequency domain data 41 ', 42 ', 43 '.
Among Fig. 5-2 originally phase place be 0 ° frequency domain data 44,46, or intensity is zero frequency domain data 45 (not shown), then need not change equally.Just, second frequency band 40 of Fig. 5-2 is adjusted into second frequency band 40 ' of Fig. 6-2 through step 13.
In like manner all frequency domain datas in the 3rd ~ the 256 group of frequency band can be adjusted to same phase place.
This step it is noted that, the adjustment amount of obtaining may limit because of hardware-as fixing every several milliseconds of ability output once, can't export synchronously etc., and must condescend to take this post in immediate time point output.
In the above-mentioned adjustment formula [formula one], function has Δ Θ, T, and dependent variable is Δ T.But also can utilize following variation example 1,2, adopt single function controlling strain and count Δ T, though the too late present embodiment of effect, but still effect was good more in the past.
[changing example 1]-reservation function T, fixedly Δ Θ.
Suppose that all frequency domain data phase difference value Δ Θ are 2 π, with first frequency band 30 of Fig. 5-1, function T=1/F
1, all frequency domain datas 31 ~ 36 are delayed one-period T according to its frequency characteristic on the frequency band under the situation that does not change phase place, just in 12 unit distances of time shaft top offset, obtain frequency band 50 as shown in Figure 7, comprise the output time point and are respectively T
13~ T
18 Frequency domain data 51 ~ 56.Then different adjustment amounts is arranged as for other frequency bands according to its frequency characteristic.
The arrangement sequential that obtains of mode is handled through subsequent step and is outputed to electrode according to this, can simulate delaying state aforementioned as the capable ripple transmission of Fig. 1-2 human body basilar membrane.
[changing example 2]-keep function Δ Θ, directly specify characteristic frequency.
Just do not make calculating parameter with the frequency characteristic of each frequency band.Suppose that all frequency bands all (are assumed to F with the fundamental frequency of this sound
5) as standard, then the frequency domain data on all frequency bands is adjusted adjustment amount Δ T=(Δ Θ/2 π) * (1/F
5), only relevant with phase difference value Δ Θ.Therefore but the fundamental frequency of sound is relevant with tone, and on-fixed, though do not classify function as changing example 2 T, also definite value not necessarily.
In addition,, also can earlier frequency domain data be decomposed into each component of a vector, carry out displacement again, be described in more detail in following variation example 3 for reducing step 13 integral operation amount.
[change example 3]-with the component is that unit carries out the tagmeme adjustment.
Change example 3 and the main difference of present embodiment and be that step 131 and step 132 add another step 131 ': vectorial decomposition.
With Fig. 5-1 time point T
3On frequency domain data 33, with frequency domain data vector on the time shaft be decomposed into shown in Fig. 8-1 with specify phase angle Θ
0Horizontal composition 331 that equates and and Θ
0Vertical vertical composition 332.So T
3The horizontal composition 331 of frequency domain data 33 need not to adjust outgoing position on the time point, and step 132 need calculate the adjustment amount of vertical composition 332.
Thus, the Δ Θ of all frequency domain datas in [formula one] is fixed with four kinds-0 (as horizontal composition 331, frequency domain data 34,36), 0.5 π (as frequency domain data 31, vertical composition 332), π (as frequency domain data 32), and 1.5 π, just the decomposition composition of all frequency domain datas of this frequency band has only four kinds of adjustment amounts-Δ T=0,0.25T, 0.5T, and 0.75T, obtain the frequency band 301 shown in Fig. 8-2, the output point of adjusting the vertical composition 332 ' in back is at T
6, former T
1, T
231,32 same adjustment of the frequency domain data of time point become T
4, T
8The frequency domain data 31 ', 32 ' of time point.In like manner can try to achieve second ~ the 256 group of frequency band data.
In addition, this variation example 3 also can deal with at the frequency domain data of non-perpendicular time shaft.As shown in Figure 9, with T
1The frequency domain data 6 of time point can be divided into first composition 61 of parallel time shaft, and second composition 62 of vertical time axle.Again, second composition 62 can be divided into the real number composition 621 of parallel R axle, and the imaginary number composition 622 of vertical R axle.That is to say that frequency domain data 6 originally equals first composition 61 and adds real number composition 621 and imaginary number composition 622.
Cooperate and consult Figure 10-1, with the former T that is positioned at
1The real number composition 621 of time point and imaginary number composition 622, adjusting respectively according to preceding method becomes time point T
7And T
4On composition 621 ' and composition 622 '.The former T that is positioned at
161 of first compositions of time point and parallel time shaft be adjusted to time point T
4And T
7, be defined as composition 611 and composition 612.
Cooperate with reference to figure 10-2, because the composition the 611, the 612nd of parallel time shaft, represent the change amount of composition 622 ', 621 ' on time shaft, this again size just equal a chronomere, therefore composition 622 ', 621 ' translation one time unit again can be obtained output point respectively at T at last
5And T
8Composition 63,64.
To change example 3 gained sequential output to electrode again through subsequent treatment content, though inequality with present embodiment, the voice signal of last inverse conversion reduction still can be identical.
Moreover, disposal route of the present invention also can use processing shown in Figure 11-1-have the situation of complex frequency domain data at one time.This situation may be by artificially painstakingly being decomposed into many frequency domain datas with former frequency domain data vector, or real, the imaginary number of former frequency domain data is considered as a frequency domain data respectively to be obtained.Among Figure 11-1, time point T
1On frequency domain data 65 phase places be that 0.25 π, frequency domain data 66 phase places are that π, frequency domain data 67 phase places are 1.5 π.Can frequency domain data 65 directly be adjusted to T according to the processing mode of present embodiment step 13
2With T
3Between, or according to aforementioned variation example 3 processing modes, frequency domain data 65 is decomposed into 0 ° horizontal composition 651 and the vertical composition 652 of 0.5 π, frequency domain data 66,67 need not decompose, shown in Figure 11-2, vertical composition 652 and frequency domain data 66,67 are adjusted into single phase place respectively and lay respectively at T at last
4, T
7, T
10The frequency domain data 652 ', 66 ', 67 ' of time point, 651 of horizontal compositions need not be adjusted.
Hence one can see that, all frequency domain datas all optionally vector resolve into a plurality of compositions, each composition is again according to its frequency characteristic and and specify the difference between phase angle to adjust adjustment amount.
Step 14-vector stack computing.
All frequency domain datas in first, second frequency band 30,40 are single phase place after step 13 is handled, so the stack of the vector on the time shaft same point is equivalent to intensity and adds up, and obtain the frequency band 30 shown in Figure 12-1, Figure 12-2 ", 40 ".If specified non-single phase place in the step 131 is then calculated in general addition of vectors mode.
For instance, Fig. 6-1 time shaft T
4Frequency domain data 31 ' on the time point is overlapping with frequency domain data 34, intensity (length) can be added up to obtain output data 34 ' just like shown in Figure 12-1; Fig. 6-2 time shaft T
4Frequency domain data 43 ' on the time point is overlapping with frequency domain data 44, intensity (length) can be added up to obtain output data 44 ' just like shown in Figure 12-2.
Same, the variation example 3 of step 13, just time point T among Fig. 8-2
4Go up overlapping frequency domain data 34 and 31 ', time point T
6Go up overlapping frequency domain data 36 and 332 ', also step is carried out intensity and is added up according to this.
Step 15-adjustment intensity.
Consult Figure 12-1, Figure 12-2, Figure 13 simultaneously because step 13 makes many frequency domain datas at the time shaft top offset, after step 14 accumulation process, during actual operation, on the time shaft the output data intensity of right (the representative time is later) enhancing trend is arranged.Therefore the output data intensity on each frequency band must multiply by a ratio value C, the signal intensity of output is adjusted in the acceptable setting range of machine, otherwise last output current is more and more strong, can cause that the user frightens, injured, instrument damage, or the safety valve that is set by language processor 22 blocks.
At present embodiment, sometime ratio value C account form may further comprise the steps in certain frequency band, with T
4Time point illustrates:
Step 151-look for tagmeme to adjust preceding frequency domain data maximum of intensity.
Look for all frequency bands before carry out step 13, time point T
4Last frequency domain data maximum of intensity.At this hypothesis time point T
4Last frequency domain data maximum of intensity is the intensity level (also equaling the intensity level of Fig. 5-2 frequency domain data 44) of the frequency domain data 34 of first frequency band 30 among Fig. 5-1.
Step 152-the look for output data maximum of intensity after adding up.
Look for all frequency bands after carry out step 14, time point T
4Last output data maximum of intensity.At this hypothesis time point T
4Last output data maximum of intensity is the intensity level of output data 44 ' among Figure 12-2.
Step 153-ask ratio value C4.
Ratio value C
4Equal the resultant frequency domain data maximum of intensity of step 151, divided by the resultant output data maximum of intensity of step 152.With this ratio value C
4Multiply by all frequency bands at time point T
4Output data intensity.
Output data 34 ' among Figure 12-1 behind step 151 ~ 153, the output data 71 that renumbers as shown in figure 13; Output data 44 ' among Figure 12-2 is this passage T originally
4Maximum output intensity on the point, after this step ratio was adjusted, it was without the preceding maximum intensity of step 13 that output intensity is replied.
In like manner can try to achieve the output data 72 ~ 74 of other times point, obtain frequency band 70; Other frequency bands then according to this method calculate in addition.
Certainly the intensity method of adjustment is not exceeded with present embodiment, with identical motivation principle, and the multiple variation example of can deriving:
[changing routine I]
Similar present embodiment, but do not look for maximal value, be directly " to carry out the preceding frequency domain data intensity of step 13 " with the special frequency band time point, divided by " carrying out step 14 back output data intensity ", try to achieve ratio value C with this time point of frequency band
I, and make the output data of 256 frequency bands on this time point all be multiplied by this ratio value C
I
For example the fundamental frequency of this voice signal is F
5, be the important frequencies of this voice signal, can directly specify with the 5th frequency band and do to calculate the basis.The 5th frequency band T
4Output data on the point, after the ratio through changing routine I was adjusted, then the output intensity answer was without the intensity before the step 13.
[changing routine II]
For simulating at present artificial on the market electronic ear system 2 employed port numbers, and the limited number of electrodes that can implant, this changes routine is that a frequency band integration is respectively done in the unit front and back with the passage.Change routine II with regard to certain particular point in time, ratio value C
IIAsk method to comprise following thin step:
(1) 22 passages are carried out the preceding integration of step 13, just that all frequency bands in each passage are average again at the frequency domain data addition of vectors of this time point.
Suppose in aforementioned the 22 passage only to first frequency band 30, second frequency band 40 should be arranged, get data splitting 711 ~ 716 as shown in figure 14 after the addition, be defined as " passage integral data ", its intensity is divided by the frequency band number of this passage, get mean value as shown in figure 15, be considered as the average frequency domain data 721 ~ 726 of all frequency bands in this passage.,, but be far from it in fact usually because first, second frequency band 30,40 distributions of hypothesis are identical at this so the average frequency domain data 721 ~ 726 among Figure 15 are average preceding identical with the frequency domain data of first frequency band 30 or second frequency band 40.
" passage integral data " has 22 on the same time point, shows wherein the first stroke among Figure 14.
(2) look on the particular point in time maximal value of its 22 " passage integral data " intensity.
(3) the average frequency domain data of all frequency bands are carried out the sequential adjustment of step 13 and the accumulation process of step 14, respective frequencies is F
1Frequency domain data become output data 731 ~ 734 shown in Figure 16-1; Respective frequencies is F
2Frequency domain data then become output data 741 ~ 744 shown in Figure 16-2.
(4) integrate once more,, be used as " the passage output data " of this passage, have 22 at this time point just with the output data addition of vectors of all frequency bands in each passage at this time point.With the intensity of " passage output data " frequency band number divided by this passage, the mean value that obtains then is considered as the average output data of all frequency bands in this passage, for experimentally carrying out the inverse conversion recovering signal.
(5) look for the maximal value of these 22 " passage output data " intensity.
(6) obtain ratio value C
IIBe exactly " passage integral data " maximum of intensity divided by " passage output data " maximum of intensity, and make 256 frequency bands all be multiplied by this ratio value C in the average output data of this time point
II
What deserves to be mentioned is, the step (6) of above-mentioned variation example II is resultant, all frequency bands of corresponding same passage output valve after the intensity adjustment of this time point is all the same in fact, and therefore in fact running can be directly all be multiplied by ratio value C with the output intensity of 22 these time points of passage
IIThen next step (step 16) directly reaches counter electrode and exports.
In addition, because the frequency of cochlea corresponds to nonlinear system, and electronic ear system 2 implantable electrode numbers (or the port number of saying so) are limited at present, can't make between frequency band and electrode directly corresponding one to one, it is unit that therefore above-mentioned variation example II needs with the passage, and calculate the average frequency domain data and the average output data of each frequency band in each passage, cause the result of last inverse conversion reduction to produce a little error.If implant electrode quantity is improved, make port number near frequency band number, then do not need the process of equalization, its output and inverse conversion reduction result will be near desirable.
[changing routine III]
The routine II of similar variation, all passages carry out integrative actions such as step (1), (3), (4) equally, but do not look for maximal value, but directly specify special modality to calculate ratio value C
III
For example the fundamental frequency of this voice signal is F
5, belong to the 21 passage, therefore directly appointment obtains ratio value C with the intensity of " the passage integral data " of the 21 passage intensity divided by same passage " passage output data "
III, and make 256 frequency bands all be multiplied by this ratio value C in the average output data of this time point
III
Step 16-be passed to counter electrode.
With the resultant 256 groups of frequency bands of above-mentioned steps, according to its representative frequency range and institute's respective channel thereof, after the sound intensity is changed by force electromagnetism, adjusted parameter, coding according to the data table of comparisons (mapping (enum) data), find counter electrode, and output synchronously.Wherein mapping (enum) data is designed according to user's auditory cell and behaviour in service.
What deserves to be mentioned is, the vector stack of step 14, the resize ratio of step 15 also can be skipped, before directly treatment step such as changing by force to electromagnetism, obtain the intensity and the phase place total value of each frequency domain data on this time point or composition with the mathematical operation formula, find respective channel adjustment output again, but its technological essence with the frequency domain data combination is still identical with the present invention.
Belong to the 22 passage of low frequency in first group of frequency band 30 of present embodiment hypothesis Fig. 5-1,5-2, the frequency range of second group of frequency band 40, can be after conversion is adjusted by planting in electrode output near the cochlea tip, each electrode is with similar normal cochlea transmittance process, the time point of the capable ripple delay phenomenon of basilar memebrane stimulates cochlea, and is passed to brain through auditory nerve.The people knows from experience all frequency domain datas that will pass on voluntarily and carries out inverse conversion, restores the original sound signal.
Cooperating and consult Fig. 3-2, experimentally, is to carry out contrary fast fourier transform (IFFT), and reduction obtains shown in Fig. 3-2, extremely near original sound signal and clear and legible output signal.
In sum, signal processing method of the present invention is applied in artificial electronic ear system 2, do not change all features of original sound as far as possible, verily transmit all fundamental tones, the former feature of overtone and noise (Stochastic Resonance Phenomenon that increases hearing is reappeared) etc., the signal of giving up nonzero phase with prior art maybe all is considered as all phase places the signal processing method comparison that zero phase is exported, the present invention exports the result can transmit the phase propetry of each composition of signal under its frequency with the clue of time, and will more correct acoustic information reach the corresponding electrode of artificial electronic ear, can reach purpose of the present invention really.
Show through experimental result, the present invention and the disposal route comparison of only transmitting signal intensity, casting out phase propetry, under testee and evaluator's double blinding, (for example: the strength) test of sampling choose all the two speech of 20 national languages handling through above-mentioned two kinds of methods with random number, all passages are with same analysis and SR, and the accuracy that obtains as shown in Table 1.
Table one
By above confirmation, the present invention exports the result and ideal is very approaching, and by the lifting of port number, analysis and SR, can obtain higher accuracy, reaches in the signal transduction process the fully purpose of compression, the complete preservation of signal characteristic.
Consult Fig. 1-1, Fig. 2 and Figure 17 simultaneously, it is the speech processor 22 that is applied to artificial electronic ear system 2 that signal processing apparatus of the present invention 8 cooperates first preferred embodiment, it comprises acquisition unit 81, converting unit 82, tagmeme adjustment unit 83, the assembled unit 84 in order to execution in step 14 in order to execution in step 13 in order to execution in step 12 in order to execution in step 11 that links to each other in regular turn, and an intensity adjustment unit 85 in order to execution in step 15.
As shown in figure 18, the present invention's second preferred embodiment is the signal processing that is applied in artificial electronic eyes system 9.Artificial electronic eyes system 9 comprises that mainly a dress is worn on eyes outside (for example being placed on the glasses) or the artificial lens in order to the microcamera 91 that captures a signal of video signal 90, links to each other with microcamera 91 and in order to handle the image processor 92 of signal of video signal 90, reach the array electrode 93 on the implant into body retina.Former electrodes 93 implant sites do not exceed with retina, and are implantable yet
In optic nerve or brain visual area.
Microcamera 91 converts the two-dimensional pixel of figure to the two-dimentional electric pulse of plane distribution, it is exactly signal of video signal 90, after image processor 92 numerical codings, be passed to electrode 93, being converted to by electrode 93 directly stimulates the retina cell behind the electric wave, see through optic nerve and be passed to brain, make PVR and total blindness's patient's foundation or recovery eyesight.
When the inventive method is applied in artificial electronic eyes system, be step 11 ~ 16 of carrying out aforementioned first preferred embodiment equally, content difference changes the bidimensional image signal except handled signal into by the sequence voice signal, is that mainly definition-the time shaft of variable axle changes spatial axes into.
Present embodiment illustrates with 1000 * 1000 pixel graphics, this figure has 1000 row, each shows 1000 points, suppose that every bit all makes the FFT of one time 256 number of samples, then can obtain 256 spatial frequency band, therefore each spatial frequency band has 1000 frequency domain datas, and just each shows 1000 * 256 frequency domain datas.
The electrode implanting state that present embodiment is desirable is to comply with the corresponding spatial frequency of each electrode with 1000 * 1000 * 256 array electrode 93, plants on the amphiblestroid correspondence position of human body.But because at present about the position of human body retina, optic nerve or brain visual area and the corresponding relation between spatial frequency, the clear and definite research conclusion of Shang Weiyou, thereby be to illustrate with 1000 * 1000 * 1 array electrode 93 at this, but not as limit.Wherein the amphiblestroid ad-hoc location of each electrode implant into body respectively has its corresponding spatial frequency, receives at least one frequency band.
After frequency domain data in each spatial frequency band utilizes the inventive method to adjust space tagmeme processing, all frequency bands that correspond to same electrode are added up in the output data of the same space point, and after, the data table of comparisons strong to electromagnetism through the sound intensity etc. handle, output to this electrode.Last brain can be equal to the inverse conversion action of IFFT according to the spatial frequency output data of position on each spatial point of spatial axes (locus of pixel), and reduction obtains phase place and intensity without the former visual signal that changes.
At step 12 place, because previous research inference, the spatial frequency of corresponding human vision should be linear distribution, therefore before present embodiment whenever carries out FFT, can utilize form function (window function) that signal of video signal 80 is handled, make signal concentrated more, correct.
In addition, the present invention is applied in artificial electronic eyes system also can handle dynamic image certainly, above-mentioned figure need only be used as the figure on the time frame (frame), for the locational frequency domain data of same pixel space, its phase place and intensity level can change in time.For instance, when the dynamic image per second produces 60 frames, be equivalent to the 60Hz sampling rate, just the locational frequency domain data of each pixel space also can show on the time shaft, and therefore also can use the present invention does the adjustment of time frequency band on time shaft.
Third embodiment of the invention is to be applied in OFDM (OFDM) modulating system, and can spread to the system that other need the information of transmitting with various phase places originally thus, and the phase propetry that makes signal can be held more information after integrating compression, improve transfer efficiency.
As shown in figure 19, the signal of OFDM is made up of the subcarrier of many complex signal representatives, and these complex signals are to be converted through quadrature amplitude modulation (qam) by serial data.Each subcarrier shows on frequency domain like sine function, arranges with other subcarrier quadrature bunchiness.When the transmission end these complex signals are propagated after IFFT becomes an analog electrical magnetic wave and raising frequency, receiving end can carry out reversal procedure, obtains original signal through FFT.
When the inventive method is applied to OFDM, it is direct step 12 in aforementioned first preferred embodiment, above-mentioned complex signal is arranged on the time shaft of each frequency band, carries out step 13 ~ 16 of aforementioned first preferred embodiment then, and the appointment phase angle of thin step 131 is two complementary phase angles.
Shown in Figure 20-1, Figure 20-2, Figure 20-3, Figure 20-4, first group of frequency domain data that the complex signal of demonstration four subcarriers forms on time shaft among the figure, the frequency domain data on each time point respectively has its intensity and phase propetry.Shown in Figure 21-1, Figure 21-2, Figure 21-3, Figure 21-4, import independently second group of frequency domain data of one and first group of frequency domain data in addition, be that the complex signal by four subcarriers forms on time shaft equally.
Cooperate with reference to figure 22-1, Figure 22-2, Figure 22-3, Figure 22-4, by the inventive method, can be on each time point and present first group of frequency domain data of multiple phase place and intensity with one's own department or unit, Figure 20-1 ~ 4 Central Plains, to specify 0 ° in phase angle and π to be converted to the two phase shown in Figure 22-1 ~ 4, the output data of multiple intensity, show among the figure that phase place is 0 °, π, the just output data on the number plane.
In like manner, second group of frequency domain data shown in Figure 21-1 ~ 4, to specify phase angle 0.5 π, 1.5 π also can be converted to the two phase shown in Figure 23-1 ~ 4, the output data of multiple intensity, show among the figure that phase place is 0.5 π, 1.5 π, the just output data on the imaginary plane by the present invention.
Therefore, combination transmitted after the present invention can be converted to former OFDM data stationary phase.That is to say,, add the frequency domain data (imaginary number) of Figure 23-1 ~ 4, obtain new, a complete combined frequency domain data, carry out utilizing original ofdm system to transmit behind the IFFT the frequency domain data (real number) of Figure 22-1 ~ 4 of each each frequency band of time point.
As for using OFDM receiving end of the present invention, then be that the signal that will receive carries out FFT, can obtain real part and be the imaginary number output data among Figure 23-1 ~ 4 for the real number output data among Figure 22-1 ~ 4, imaginary part.The real number output data separated with the imaginary number output data carry out IFFT, get the real number ordered series of numbers, FFT again, then Figure 22-1 ~ 4 is reducible, and to be that Figure 20-1 ~ 4, Figure 23-1 ~ 4 are reducible be Figure 21-1 ~ 4, obtains the result almost completely identical with original input signal.
Therefore in sum, the present invention is applied in ofdm system, can additionally transmit another group independent data, can realize achievements such as two times of fast ADSL, two times of fast wireless network cards, two times of speed digitals be audio-visual.
Claims (41)
1, a kind of signal processing method is handled at a plurality of frequency domain datas in the identical frequency band scope, and each frequency domain data includes intensity and phase propetry, it is characterized in that, this method comprises the following steps:
These frequency domain datas are arranged on the variable axle;
B. each frequency domain data is a phase parameter with its phase propetry, in order to adjust the position of this frequency domain data on this variable axle; And
C. the frequency domain data combination that will adjust to same position on the variable axle is for output.
2, signal processing method as claimed in claim 1 is characterized in that: described this frequency domain data of step B is a frequency parameter with the frequency characteristic of its place frequency band also, in order to the position of this frequency domain data of common adjustment on this variable axle.
3, signal processing method as claimed in claim 2 is characterized in that: the frequency characteristic of this frequency band is meant the centre frequency of its frequency band range.
4, signal processing method as claimed in claim 2 is characterized in that: the adjustment amount of described this frequency domain data position of step B is under this frequency characteristic condition, during this frequency domain data phase change, and corresponding variable quantity on this variable axle.
5, signal processing method as claimed in claim 2, it is characterized in that: step B is further comprising the steps of: (B1) set at least one appointment phase angle; (B2) calculate equidirectional lowest difference between this phase parameter and appointment phase angle, obtain a phase difference value; (B3) calculate the adjustment amount of this frequency domain data on this variable axle, equal this phase difference value, be multiplied by the inverse of this frequency parameter again divided by 2 π; The phase place of this frequency domain data is specified phase angle for this after the adjustment.
6, signal processing method as claimed in claim 5 is characterized in that: on the same position of described this variable axle of steps A, be arranged with a plurality of frequency domain datas.
7, as claim 5 or 6 described signal processing methods, it is characterized in that: step B more comprises the thin step (B1-1) between step (B1) and the step (B2), divide first composition that solves a parallel variable axle with each frequency domain data vector, and second composition of a vertical variable axle, define this first composition and be the change amount of this frequency domain data on the variable axle; The last adjustment amount of this frequency domain data is that step (B3) value of obtaining is added this change amount.
8, as claim 5 or 6 described signal processing methods, it is characterized in that: step B also comprises the step (B1-2) between thin step (B1) and the step (B2), divide the parallel composition that solves parallel this appointment phase angle with each frequency domain data vector, and one vertically is somebody's turn to do the vertical composition of specifying phase angle; Step (B2) is to calculate this parallel, vertical composition and specify equidirectional minimal difference between phase angle.
9, as claim 4 or 5 described signal processing methods, it is characterized in that: on this variable axle a plurality of output points are arranged, the adjustment amount of this frequency domain data on this variable axle is to yield in immediate output point.
10, signal processing method as claimed in claim 1, it is characterized in that: the position adjustment amount of described this frequency domain data of step B, be one according to this frequency domain data under the position appointed frequency condition on the variable axle, during this frequency domain data phase change, corresponding variable quantity on this variable axle.
11, signal processing method as claimed in claim 1, it is characterized in that: step B is further comprising the steps of: (B1) set at least one appointment phase angle; (B2) calculate equidirectional lowest difference between this frequency domain data and appointment phase angle, obtain a phase difference value; (B3) calculate the adjustment amount of this frequency domain data on the variable axle, equal this phase difference value, be multiplied by one again according to the specified periodic quantity in the position of this frequency domain data on the variable axle divided by 2 π.
12, signal processing method as claimed in claim 1 is characterized in that: the array mode of step C frequency domain data is an addition of vectors.
13, signal processing method as claimed in claim 1, it is characterized in that: this method also comprises the step D after the step C, with each locational combination back frequency domain data of variable axle, under the condition that does not influence the inverse conversion reduction result, be multiplied by one in order to control output intensity in the ratio value of setting range, as the output data on this variable shaft position.
14, signal processing method as claimed in claim 1 is characterized in that: its processing signals is a tandem voice signal, and the variable axle in the steps A is a time shaft.
15, signal processing method as claimed in claim 14 is characterized in that: its processing signals is by fourier transform, this voice signal is converted to these frequency domain datas that can show its frequency and intensity, phase propetry.
16, signal processing method as claimed in claim 14 is characterized in that: its processing signals is by Hilbert conversion, this voice signal is converted to these frequency domain datas that can show its frequency and intensity, phase propetry.
17, signal processing method as claimed in claim 1 is characterized in that: its processing signals is the bidimensional image signal of a plane distribution, and the variable axle in the steps A is a spatial axes.
18, signal processing method as claimed in claim 17 is characterized in that: its processing signals is by fourier transform, this signal of video signal is converted to these frequency domain datas that can show its spatial frequency and intensity, phase propetry.
19, signal processing method as claimed in claim 18 is characterized in that: before whenever carrying out fourier transform, utilize the form function that this signal of video signal is handled.
20, signal processing method as claimed in claim 17 is characterized in that: its processing signals is by Hilbert conversion, this signal of video signal is converted to these frequency domain datas that can show its spatial frequency and intensity, phase propetry.
21, a kind of signal processing method, handle at a string column signal, this signal comprises and a plurality ofly can be distinctly adheres to one of them frequency domain data of a plurality of frequency bands separately according to its frequency, and each frequency band has its corresponding frequency range, and each frequency domain data includes intensity and phase propetry; It is characterized in that this method comprises the following step:
The frequency domain data that belongs to a frequency band together is arranged on the variable axle;
B. each frequency domain data is a parameter with the frequency characteristic of its place frequency band, adjusts the position of this frequency domain data on this variable axle.
22, a kind of signal processing apparatus, in order to handle a plurality of frequency domain datas in the identical frequency band scope, each frequency domain data includes intensity and phase propetry, it is characterized in that this device comprises:
One converting unit, with so that these frequency domain datas be arranged on the variable axle;
One tagmeme adjustment unit is connected with this converting unit, in order to be a phase parameter with each frequency domain data with its phase propetry, adjusts the position of this frequency domain data on this variable axle; And
One assembled unit is connected with this tagmeme adjustment unit, makes up for output in order to the frequency domain data that will adjust to same position on the variable axle.
23, signal processing apparatus as claimed in claim 22 is characterized in that: this tagmeme adjustment unit is a frequency parameter with the frequency characteristic of this frequency domain data place frequency band more, adjusts the position of this frequency domain data on this variable axle.
24, signal processing apparatus as claimed in claim 23 is characterized in that: the frequency characteristic of this frequency band is meant the centre frequency of its frequency band range.
25, signal processing apparatus as claimed in claim 23 is characterized in that: this tagmeme adjustment unit is under this frequency characteristic condition for the adjustment amount of this frequency domain data position, during this frequency domain data phase change, and corresponding variable quantity on this variable axle.
26, signal processing apparatus as claimed in claim 23, it is characterized in that: this tagmeme adjustment unit is to set at least one appointment phase angle earlier, calculate this phase parameter again and specify equidirectional lowest difference between phase angle, obtain a phase difference value, calculate the adjustment amount of this frequency domain data on the variable axle at last, this adjustment amount equals this phase difference value divided by 2 π, is multiplied by the inverse of this frequency parameter again; The phase place of this frequency domain data is specified phase angle for this after the adjustment.
27, signal processing apparatus as claimed in claim 26 is characterized in that: this converting unit is arranged in a plurality of frequency domain datas on the same position of this variable axle jointly.
28, as claim 26 or 27 described signal processing apparatus, it is characterized in that: this tagmeme adjustment unit is to divide first composition that solves a parallel variable axle with each frequency domain data vector, and second composition of a vertical variable axle, and this tagmeme adjustment unit adjustment amount of adjusting this frequency domain data should be added the amount of this first composition.
29, as claim 26 or 27 described signal processing apparatus, it is characterized in that: this tagmeme adjustment unit divides the parallel composition that solves parallel this appointment phase angle with each frequency domain data vector, and one vertically is somebody's turn to do the vertical composition of specifying phase angle; And this tagmeme adjustment unit calculates this parallel, vertical composition respectively and specifies equidirectional minimal difference between phase angle.
30, as claim 25 or 26 described signal processing apparatus, it is characterized in that: on this variable axle a plurality of output points are arranged, this tagmeme adjustment unit is obtained the adjustment amount of this frequency domain data on this variable axle, is to yield in immediate output point.
31, signal processing apparatus as claimed in claim 22, it is characterized in that: this tagmeme adjustment unit is adjusted the adjustment amount of this frequency domain data position, be one according to this frequency domain data under the position appointed frequency condition on the variable axle, during this frequency domain data phase change, corresponding variable quantity on this variable axle.
32, signal processing apparatus as claimed in claim 22, it is characterized in that: this tagmeme adjustment unit is to set at least one appointment phase angle earlier, calculate this frequency domain data again and specify equidirectional lowest difference between phase angle, obtain a phase difference value, calculate the adjustment amount of this frequency domain data on the variable axle at last, this adjustment amount equals this phase difference value divided by 2 π, is multiplied by one again according to the specified periodic quantity in the position of this frequency domain data on the variable axle.
33, signal processing apparatus as claimed in claim 22 is characterized in that: this assembled unit is to utilize addition of vectors that frequency domain data is made up.
34, signal processing apparatus as claimed in claim 22, it is characterized in that: this device more comprises an intensity adjustment unit that is connected with this assembled unit, in order to each locational combination back frequency domain data of variable axle, under the condition that does not influence the inverse conversion reduction result, be multiplied by one in order to control output intensity in the ratio value of setting range, as the output data on this variable shaft position.
35, signal processing apparatus as claimed in claim 22 is characterized in that: its processing signals is a tandem voice signal, and described variable axle is a time shaft.
36, signal processing apparatus as claimed in claim 35 is characterized in that: this converting unit is to utilize fourier transform, this voice signal is converted to these frequency domain datas that can show its frequency and intensity, phase propetry.
37, signal processing apparatus as claimed in claim 35 is characterized in that: this converting unit is to utilize Hilbert conversion, this voice signal is converted to these frequency domain datas that can show its frequency and intensity, phase propetry.
38, signal processing apparatus as claimed in claim 22 is characterized in that: its processing signals is the bidimensional image signal of a plane distribution, and described variable axle is a spatial axes.
39, signal processing apparatus as claimed in claim 38 is characterized in that: this converting unit is to utilize fourier transform, this signal of video signal is converted to these frequency domain datas that can show its spatial frequency and intensity, phase propetry.
40, signal processing apparatus as claimed in claim 39 is characterized in that: this converting unit utilizes the form function that this signal of video signal is handled before whenever carrying out fourier transform.
41, signal processing apparatus as claimed in claim 38 is characterized in that: this converting unit is to utilize Hilbert conversion, this signal of video signal is converted to these frequency domain datas that can show its spatial frequency and intensity, phase propetry.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNB2004100489673A CN100514447C (en) | 2004-06-15 | 2004-06-15 | Signal proces and its modular set |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNB2004100489673A CN100514447C (en) | 2004-06-15 | 2004-06-15 | Signal proces and its modular set |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN1713274A CN1713274A (en) | 2005-12-28 |
| CN100514447C true CN100514447C (en) | 2009-07-15 |
Family
ID=35718853
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CNB2004100489673A Expired - Fee Related CN100514447C (en) | 2004-06-15 | 2004-06-15 | Signal proces and its modular set |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN100514447C (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| TWI761671B (en) | 2019-04-02 | 2022-04-21 | 緯創資通股份有限公司 | Living body detection method and living body detection system |
-
2004
- 2004-06-15 CN CNB2004100489673A patent/CN100514447C/en not_active Expired - Fee Related
Also Published As
| Publication number | Publication date |
|---|---|
| CN1713274A (en) | 2005-12-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103140260B (en) | Cochlear implant stimulation with low frequency channel privilege | |
| CN101642399B (en) | Artificial cochlea speech processing method based on frequency modulation information and artificial cochlea speech processor | |
| CN100502819C (en) | Artificial cochlea manufacture method suitable for Chinese voice coding strategy | |
| CN106782565A (en) | A kind of vocal print feature recognition methods and system | |
| CN109545233A (en) | System and method for handling audio signal to be reset on audio devices | |
| CN105489227A (en) | Hearing device comprising a low-latency sound source separation unit | |
| CN102958561A (en) | Envelope specific stimulus timing | |
| AU2005202837A1 (en) | Selective resolution speech processing | |
| CN101695148A (en) | Multi-channel wide dynamic range compressing system for digital hearing aid | |
| AU2004208738B2 (en) | Signal processing method and module | |
| CN104464741B (en) | A kind of audio signal turns the method and system of visual color information | |
| CN101646123A (en) | Filter bank simulating auditory perception model | |
| TWI398243B (en) | Electrode simulating method and system thereof | |
| CN100514447C (en) | Signal proces and its modular set | |
| CN102525734A (en) | Electronic cochlea virtual channel signal processing system based on peak value selection and method | |
| EP2732641B1 (en) | Test methods for cochlear implant stimulation strategies | |
| Huang et al. | Perceptual fusion tendency of speech sounds | |
| US20090281810A1 (en) | System and method for visually presenting audio signals | |
| CN111508500B (en) | Voice emotion recognition method, system, device and storage medium | |
| Hu et al. | A model framework for simulating spatial hearing of bilateral cochlear implant users | |
| EP1635609B1 (en) | Signal processing method and apparatus | |
| Nie et al. | Implementation of CIS speech processing strategy for cochlear implants by using wavelet transform | |
| Liu et al. | A novel speech coding algorithm for cochlear implants | |
| Zotkin et al. | Neuromimetic sound representation for percept detection and manipulation | |
| CN108665904A (en) | A kind of external speech signal processing system of cochlear implant and cochlear implant |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20090715 Termination date: 20170615 |