CN104363003A - Filtering method and device - Google Patents
Filtering method and device Download PDFInfo
- Publication number
- CN104363003A CN104363003A CN201410584151.6A CN201410584151A CN104363003A CN 104363003 A CN104363003 A CN 104363003A CN 201410584151 A CN201410584151 A CN 201410584151A CN 104363003 A CN104363003 A CN 104363003A
- Authority
- CN
- China
- Prior art keywords
- operator
- group
- individual
- data
- input
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 84
- 238000001914 filtration Methods 0.000 title claims abstract description 41
- 238000004364 calculation method Methods 0.000 claims abstract description 106
- 230000008569 process Effects 0.000 claims abstract description 46
- 238000005498 polishing Methods 0.000 claims description 19
- 230000007812 deficiency Effects 0.000 claims description 13
- 230000004044 response Effects 0.000 claims description 7
- 238000012545 processing Methods 0.000 abstract description 13
- 230000000694 effects Effects 0.000 abstract description 11
- 239000000284 extract Substances 0.000 description 7
- 238000010586 diagram Methods 0.000 description 6
- 230000005236 sound signal Effects 0.000 description 4
- 238000006243 chemical reaction Methods 0.000 description 3
- 238000005070 sampling Methods 0.000 description 3
- 230000006870 function Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000009977 dual effect Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 238000012163 sequencing technique Methods 0.000 description 1
Landscapes
- Telephonic Communication Services (AREA)
Abstract
The invention discloses a filtering method and a filtering device, and belongs to the digital signal processing field. The filtering method includes: receiving a sampled signal to be processed; performing convolution calculation on the sampled signal according to calculation branches respectively corresponding to preset N groups of operators, wherein the N groups of the operators are obtained by sequentially cutting filtering operators of the finite length, the length of each group of the operators meets power of 2, the power is a positive integer larger than or equal to 1, and N is larger than or equal to 2; merging calculation results of the calculation branches respectively corresponding to the N groups of the operators so as to obtain a merged calculation result; outputting the merged calculation result as a processing result of the sampled signal. According to the filtering method and the filtering device, the operators of the convolution calculation are divided into the N groups, the convolution calculation is performed between each operator and the sampled signal, and then the convolution calculation results are merged, a long convolution processing process of the sampled signal is divided into N short convolution processing processes which are synchronously performed, and therefore filtering processing delay involved in the convolution calculation is reduced, and filtering effects are improved.
Description
Technical field
The present invention relates to digital processing field, particularly a kind of filtering method and device.
Background technology
There is the filtering of limit for length's unit impulse response (English: Finite Impulse Response, abbreviation: FIR) filtering is a kind of important digital signal processing method, is generally used for the fields such as audio frequency and video process, speech processes, pattern matching, energy conversion and electrical network.
In existing FIR filtering method, first determine FIR filter operator according to the application demand of reality, i.e. n rank FIR filter factor [a
0, a
1a
n-1], then by n sampled signal every in sampled signal after Fourier transform processing, do convolutional calculation with this n rank filter operator, the result of calculation of output can obtain filter result through inverse Fourier transform process again.In above-mentioned convolutional calculation process, the calculating duration of convolutional calculation is relevant to the length of FIR filter operator, the length of operator is longer, the computing time of convolutional calculation is longer, the delay of filtering is also higher, and the filter effect reaching expectation often needs longer FIR filter operator, thus have a strong impact on the effect of signal transacting.
Summary of the invention
Longer and cause the delay of filtering higher in order to solve FIR filter operator in prior art, thus affect the problem of the effect of signal transacting, embodiments provide a kind of filtering method and device.Described technical scheme is as follows:
First aspect, provides a kind of filtering method, and described method comprises:
Receive pending sampled signal;
By the N group operator each self-corresponding Branch Computed pre-set, convolutional calculation is carried out to described sampled signal, described N group operator is obtain by cutting successively the filter operator of finite length, the length often organizing operator meet 2 power side and power be more than or equal to 1 positive integer, N >=2;
The result of calculation of each self-corresponding Branch Computed of described N group operator is merged;
Result of calculation after merging is exported as the result to described sampled signal.
Optionally, the described each self-corresponding Branch Computed of N group operator by pre-setting carries out convolutional calculation to described sampled signal, comprising:
When to receive at first and not reached n by the number of the sampled signal of Branch Computed process corresponding to p group operator
ptime, to described n
pindividual sampled signal carries out Fourier transform, obtains n
pindividual Fourier transform data, n
pfor the number of described p group operator, 1≤p≤N;
By described n
pindividual Fourier transform data and described p group operator carry out convolutional calculation, obtain n
pindividual convolutional calculation data;
To described n
pindividual convolutional calculation data carry out inverse Fourier transform, obtain n
pindividual inverse Fourier transform data;
By described n
pindividual inverse Fourier transform data export the result of calculation for Branch Computed corresponding to described p group operator.
Optionally, described N group operator is to there being respective First Input First Output; The described result of calculation to each self-corresponding Branch Computed of described N group operator merges, and comprising:
For the 1st group of operator, by the n that the described 1st group of Branch Computed that operator is corresponding exports
1individual data and the n extracted from First Input First Output corresponding to N group operator
1individual data are added, and will be added the n obtained
1the described 1st group of First Input First Output that operator is corresponding of individual data input, the data deficiencies n extracted in the First Input First Output corresponding from described N group operator
1time individual, not enough part with 0 polishing, n
1for the number of described 1st group of operator;
For q group operator, by the n that Branch Computed corresponding for described q group operator exports
qindividual data and the n extracted from First Input First Output corresponding to q-1 group operator
qindividual data are added, and will be added the n obtained
qthe First Input First Output that the described q group operator of individual data input is corresponding, the data deficiencies n extracted in the First Input First Output corresponding from described q-1 group operator
qtime individual, not enough part with 0 polishing, n
qfor the number of described q group operator, 2≤q≤N.
Optionally, described is the result to described sampled signal by the result of calculation output after merging, comprising:
Data in the described 1st group of First Input First Output that operator is corresponding of input are exported as described result.
Optionally, the filter operator of described finite length is the filter coefficient having limit for length's unit impulse response filter.
Second aspect, provides a kind of filter, and described device comprises:
Receiver module, for receiving pending sampled signal;
Computing module, for the N group operator each self-corresponding Branch Computed by pre-setting, convolutional calculation is carried out to described sampled signal, described N group operator is obtain by cutting successively the filter operator of finite length, the length often organizing operator meet 2 power side and power be more than or equal to 1 positive integer, N >=2;
Merge module, for merging the result of calculation of each self-corresponding Branch Computed of described N group operator;
Output module, for exporting the result of calculation after merging as the result to described sampled signal.
Optionally, described computing module, comprising:
Converter unit, for ought receiving at first and not reached n by the number of the sampled signal of Branch Computed process corresponding to p group operator
ptime, to described n
pindividual sampled signal carries out Fourier transform, obtains n
pindividual Fourier transform data, n
pfor the number of described p group operator, 1≤p≤N;
Convolutional calculation unit, for by described n
pindividual Fourier transform data and described p group operator carry out convolutional calculation, obtain n
pindividual convolutional calculation data;
Inverse transform unit, for described n
pindividual convolutional calculation data carry out inverse Fourier transform, obtain n
pindividual inverse Fourier transform data;
Output unit, for by described n
pindividual inverse Fourier transform data export the result of calculation for Branch Computed corresponding to described p group operator.
Optionally, described merging module, comprising:
First sum unit, for the n exported by the 1st group of Branch Computed that operator is corresponding
1individual data and the n extracted from First Input First Output corresponding to N group operator
1individual data are sued for peace, by the n that summation obtains
1the described 1st group of First Input First Output that operator is corresponding of individual data input, the data deficiencies n extracted in the First Input First Output corresponding from described N group operator
1time individual, not enough part with 0 polishing, n
1for the number of described 1st group of operator;
Second sum unit, for the n exported by Branch Computed corresponding for described q group operator
qindividual data and the n extracted from First Input First Output corresponding to q-1 group operator
qindividual data are sued for peace, by the n that summation obtains
qthe First Input First Output that the described q group operator of individual data input is corresponding, the data deficiencies n extracted in the First Input First Output corresponding from described q-1 group operator
qtime individual, not enough part with 0 polishing, n
qfor the number of described q group operator, 2≤q≤N;
Wherein, described N group operator is to there being respective First Input First Output.
Optionally, described output module, for exporting the data in the described 1st group of First Input First Output that operator is corresponding of input as described result.
Optionally, the filter operator of described finite length is the filter coefficient having limit for length's unit impulse response filter.
The beneficial effect that the technical scheme that the embodiment of the present invention provides is brought is:
By the operator of convolutional calculation is divided into N group, and carry out convolutional calculation with sampled signal respectively, again the result of convolutional calculation is merged, by one, N number of short deconvolution process synchronously carried out is divided into the long deconvolution process of sampled signal, thus reduce the delay relating to the filtering process of convolutional calculation, improve filter effect.
Accompanying drawing explanation
In order to be illustrated more clearly in the technical scheme in the embodiment of the present invention, below the accompanying drawing used required in describing embodiment is briefly described, apparently, accompanying drawing in the following describes is only some embodiments of the present invention, for those of ordinary skill in the art, under the prerequisite not paying creative work, other accompanying drawing can also be obtained according to these accompanying drawings.
Fig. 1 is the method flow diagram of the filtering method that one embodiment of the invention provides;
Fig. 2 is the method flow diagram of the filtering method that another embodiment of the present invention provides;
Fig. 3 is the signal processing flow schematic diagram that another embodiment of the present invention provides;
Fig. 4 is the structure drawing of device of the filter that one embodiment of the invention provides;
Fig. 5 is the structure drawing of device of the filter that one embodiment of the invention provides;
Fig. 6 is the structural representation of the filter that one embodiment of the invention provides.
Embodiment
For making the object, technical solutions and advantages of the present invention clearly, below in conjunction with accompanying drawing, embodiment of the present invention is described further in detail.
Please refer to Fig. 1, it illustrates the method flow diagram of the filtering method that one embodiment of the invention provides.The method may be used for carrying out filtering process to the digital sampled signal of audio frequency or video in FIR filter.The method can comprise:
Step 102, receives pending sampled signal;
Step 104, by the N group operator each self-corresponding Branch Computed pre-set, convolutional calculation is carried out to this sampled signal, this N group operator is obtain by cutting successively the filter operator of finite length, the length often organizing operator meet 2 power side and power be more than or equal to 1 positive integer, N >=2;
Step 106, merges the result of calculation of each self-corresponding Branch Computed of this N group operator;
Step 108, exports the result of calculation after merging as the result to this sampled signal.
Be used for FIR filtering with said method and be calculated as example, in advance each FIR filter operator can be divided into N group, each group filter operator all meets the power side of 2, the part of the power side less than 2 is by 0 element polishing, obtain above-mentioned N group operator thus, after carrying out convolutional calculation with sampled signal respectively by N group operator, convolution results is merged.Because the length often organizing operator reduces relative to the length of FIR filter factor, and sampled signal carries out convolution respectively process with N group operator can the carrying out of near-synchronous, one is equivalent to be divided into N number of short deconvolution process synchronously carried out to the long deconvolution process of sampled signal, therefore can reduce the delay of FIR filtering greatly, improve filter effect.
Wherein, the Branch Computed that each group operator is corresponding can pass through software simulating, also can pass through hardware implementing.When by software simulating, this Branch Computed can be a calculation procedure; When by hardware implementing, the computing circuit that this Branch Computed can be made up of several logical blocks.
In sum, the method that the embodiment of the present invention provides, by the operator of convolutional calculation is divided into N group, and carry out convolutional calculation with sampled signal respectively, again the result of convolutional calculation is merged, by one, N number of short deconvolution process synchronously carried out is divided into the long deconvolution process of sampled signal, thus reduces the delay relating to the filtering process of convolutional calculation, improve filter effect.
Please refer to Fig. 2, it illustrates the method flow diagram of the filtering method that another embodiment of the present invention provides.The method may be used for carrying out filtering process to the digital sampled signal of audio frequency or video in FIR filter.The method can comprise:
Step 202, receives pending sampled signal;
Wherein, be used for for this FIR filter the audio signal audio signal of 5.1 sound channels being converted into dual track, this sampled signal can be the digital signal obtained after sampling according to predetermined period to the audio signal of 5.1 sound channels.
Step 204, by the N group operator each self-corresponding Branch Computed pre-set, convolutional calculation is carried out to this sampled signal, this N group operator is obtain by cutting successively the filter operator of finite length, the length often organizing operator meet 2 power side and power be more than or equal to 1 positive integer, N >=2;
Wherein, this N group operator can carry out grouping acquisition to the filter operator of known FIR filtering in advance by technical staff.This known FIR filter operator is the filter operator that technical staff pre-determines according to the actual demand of FIR filtering.Concrete, the filter operator of FIR filtering is cut into N group according to filter order order from low to high by technical staff successively, and the length often organizing operator is the power side of 2, and power be more than or equal to 1 positive integer (namely each group filter operator length meet 2,4,8,16,32 ...), to facilitate follow-up merging treatment.Wherein, when the length of last group filter operator does not meet the power side of 2,0 element polishing can be passed through.
Concrete, carrying out the process of convolutional calculation by each self-corresponding Branch Computed of N group operator that pre-sets to this sampled signal can be as follows:
1) for the Branch Computed that p group operator is corresponding, receive at first when FIR filter and do not reached n by the number of the sampled signal of Branch Computed process corresponding to p group operator
ptime, to this n
pindividual sampled signal carries out Fourier transform, obtains n
pindividual Fourier transform data, n
pbe the number of p group operator, 1≤p≤N;
Wherein, each sampled signal input filter one by one, the sequencing that each computing unit inputs according to sampled signal calculates each sampled signal in batches, and the number of the sampled signal that the computing unit that p group operator is corresponding processes at every turn is identical with the number of p group operator.The Branch Computed that p group operator is corresponding extracts not by the n of this Branch Computed process at every turn
pindividual sampled signal processes, and wherein, is not greater than n by the sampled signal of this Branch Computed process when input filter
ptime individual, extract the n wherein inputted at first
pindividual sampled signal, is not greater than n by the sampled signal of this Branch Computed process when input filter
ptime individual, then wait for that subsequent sampling signal continues input, until input filter and do not reached n by the sampled signal of this Branch Computed process
pindividual.
2) by this n
pindividual Fourier transform data and this p group operator carry out convolutional calculation, obtain n
pindividual convolutional calculation data;
3) to this n
pindividual convolutional calculation data carry out inverse Fourier transform, obtain n
pindividual inverse Fourier transform data;
4) by this n
pindividual inverse Fourier transform data export the result of calculation for Branch Computed corresponding to this p group operator.
Concrete, please refer to the signal processing flow schematic diagram in the multimedia system shown in Fig. 3, wherein, the audio samples being input as 5.1 sound channels in Fig. 3, each sampled signal needs to organize operator with each respectively and carries out convolutional calculation, for the Branch Computed that each group operator is corresponding, when untreated sampled signal number is identical with the number of this group operator, namely Fourier transform is carried out to untreated sampled signal, the data obtained after conversion and this group operator are carried out convolution algorithm, again inverse Fourier transform is carried out to the data that computing obtains, the result of calculation of Branch Computed corresponding to this group operator can be obtained.
It should be noted that, the number of the sampled signal that input once inputs is identical with the number of the 1st group of operator, when one group of sampled signal of input is after the Branch Computed process that the 1st group of operator is corresponding, then inputs next group sampled signal.
Step 206, merges the result of calculation of each self-corresponding Branch Computed of this N group operator;
For the 1st group of operator, by the n that the 1st group of Branch Computed that operator is corresponding exports
1individual data and the n extracted from First Input First Output corresponding to N group operator
1individual data are added, and will be added the n obtained
1individual data input the 1st group of First Input First Output that operator is corresponding, the data deficiencies n extracted in the First Input First Output corresponding from N group operator
1time individual, not enough part is with 0 polishing;
For q group operator, by the n that Branch Computed corresponding for this q group operator exports
qindividual data and the n extracted from First Input First Output corresponding to q-1 group operator
qindividual data are added, and will be added the n obtained
qindividual data input First Input First Output corresponding to this q group operator, the data deficiencies n extracted in the First Input First Output corresponding from q-1 group operator
qtime individual, not enough part with 0 polishing, n
qbe the number of q group operator, 2≤q≤N.
Please refer to Fig. 3, wherein, each group operator is except a corresponding Branch Computed is for except convolutional calculation, also corresponding one merges branch and a first-in first-out (English: First Input First Output, abbreviation FIFO) queue, for merging the result of calculation of Branch Computed corresponding to this group operator Branch Computed corresponding with upper one group of operator.When the merging branch that each group operator is corresponding receives one group of result data of corresponding Branch Computed input, this merging branch extracts and the same number of data of this group operator from fifo queue corresponding to upper one group of operator, and the data that Branch Computed inputs are added with the data extracted from fifo queue, to the data that obtain be added stored in fifo queue corresponding to this group operator, so that the merging branch that next group operator is corresponding extracts, wherein, the number merging the data that branch's addition obtains is identical with the number of this group operator.Further, when the number of the data in the fifo queue that upper one group of operator is corresponding is less than the number of this group operator, not enough part is by 0 element polishing.It should be noted that, merging branch corresponding to the 1st group of operator extracts data from fifo queue corresponding to N group operator.
Concrete, for q group operator, the corresponding Branch Computed q of q group operator, merging branch q and fifo queue q.After Branch Computed q completes a convolutional calculation, export result of calculation [x to merging branch q
1, x2 ..., x
m], wherein, m=n
q.Merge branch q and extract n from fifo queue q-1
qindividual data [y
1, y
2..., y
m], and the data inputted by Branch Computed q merge with the data extracted, and obtain the data [x after merging
1+ y
1, x
2+ y
2..., x
m+ y
m], and by the data after merging stored in fifo queue q.
Wherein, as the data deficiencies n in fifo queue q-1
qtime individual, not enough part at end with 0 polishing.Further, when q is 1, merges branch q from fifo queue N, extract n1 data.
Wherein, the Branch Computed that each group operator is corresponding and merging branch can pass through software simulating, also can pass through hardware implementing.When by software simulating, this Branch Computed or merging branch can be calculation procedure; When by hardware implementing, the computing circuit that this Branch Computed or merging branch can be made up of several logical blocks.
Step 208, exports the result of calculation after merging as the result to this sampled signal.
Data in input the 1st group of First Input First Output that operator is corresponding export as this result by filter.
Please refer to Fig. 3, wherein, the data in fifo queue 1, except for merging except branch 2 extraction, also export the result for whole signal processing flow.
Concrete, so that the filter operator of FIR filter is divided into three groups, if A is the input of filter, B is the output of system, L1 is first paragraph FIR coefficient (length is 32), L2 is second segment FIR coefficient (length is 256 length), and L3 is the 3rd section of FIR coefficient (length is 1024).M1, M11, M2, M3 and M4 are can the fifo area (infinite in length system) of store data.MT1 and MT2 is the temporary realm (non-FIFO, infinite in length system) of store data.Its filter step can be as follows:
1) M1, M11, M2, M3 and M4 is emptied;
2) time accumulative enough 32 data of A (these data carry out sampling to audio signal to obtain sampled signal), by these 32 data stored in M1, M11 and MT1;
3) from M2, get 32 data and stored in MT2, if less than 32 data in M2, then not enough part is with 0 polishing;
4) data of 32 in MT1 and L1 are carried out convolution algorithm, operation result is added with the data of 32 in MT2, and result is stored in M3 and export to B;
5) step 2 ~ 4 are repeated, until containing 256 data in M1;
6) 256 data stored in MT1 from M1;
7) from M3, take out 256 data and stored in MT2, if M3 is less than 256 data, then not enough part is with 0 polishing;
8) data of 256 in MT1 and L2 are carried out convolution algorithm, operation result is added with the data of 32 in MT2, and result is stored in M4;
9) step 2 ~ 8 are repeated, until containing 1024 data in M11;
10) from M11,1024 data and stored in MT1 are taken out;
11) from M4, take out 1024 data and stored in MT2, if M4 is less than 1024 data, then not enough part is with 0 polishing;
12) data of 1024 in MT1 and L3 are carried out convolution algorithm, operation result is added with the data of 32 in MT2, and result is stored in M2;
13) above-mentioned steps 2 ~ 12 is repeated, until follow-up all input data processings complete.
It should be noted that, above-mentioned example is only illustrated filter operator to be divided into 3 groups, in actual applications, according to actual needs operator can be divided at least two groups, such as, 2 groups, 4 groups or 6 groups etc. can be divided into, to this, the embodiment of the present invention is not specifically limited.In the examples described above, the number of 3 groups of filter operators increases progressively successively, in actual applications, the magnitude relationship of N group operator does not limit, such as, in actual applications, can the number of N group operator be set to equal, or, the number of N group operator can be set to successively decrease successively, to this, the embodiment of the present invention is not specifically limited equally.
In addition, filtering method shown in the embodiment of the present invention is only illustrated for carrying out process by FIR filter to audio samples, in actual applications, this method can also be used for the application scenarios that video frequency signal processing, Speech processing, pattern matching, energy conversion etc. relate to FIR filtering.Further, the filtering method shown in the embodiment of the present invention can also be applied in other and relate in the filter of convolution algorithm.
In sum, the method that the embodiment of the present invention provides, by the operator of convolutional calculation is divided into N group, and carry out convolutional calculation with sampled signal respectively, again the result of convolutional calculation is merged, by one, N number of short deconvolution process synchronously carried out is divided into the long deconvolution process of sampled signal, thus reduces the delay relating to the filtering process of convolutional calculation, improve filter effect.
Please refer to Fig. 4, it illustrates the structure drawing of device of the filter that one embodiment of the invention provides.This device can be implemented as the part in FIR filter or FIR filter.This device can comprise:
Receiver module 301, for receiving pending sampled signal;
Computing module 302, for the N group operator each self-corresponding Branch Computed by pre-setting, convolutional calculation is carried out to this sampled signal, this N group operator is obtain by cutting successively the filter operator of finite length, the length often organizing operator meet 2 power side and power be more than or equal to 1 positive integer, N >=2;
Merge module 303, for merging the result of calculation of each self-corresponding Branch Computed of this N group operator;
Output module 304, for exporting the result of calculation after merging as the result to this sampled signal.
In sum, the device that the embodiment of the present invention provides, by the operator of convolutional calculation is divided into N group, and carry out convolutional calculation with sampled signal respectively, again the result of convolutional calculation is merged, by one, N number of short deconvolution process synchronously carried out is divided into the long deconvolution process of sampled signal, thus reduces the delay relating to the filtering process of convolutional calculation, improve filter effect.
Please refer to Fig. 5, it illustrates the structure drawing of device of the filter that another embodiment of the present invention provides.This device can be implemented as the part in FIR filter or FIR filter.This device can comprise:
Receiver module 301, for receiving pending sampled signal;
Computing module 302, for the N group operator each self-corresponding Branch Computed by pre-setting, convolutional calculation is carried out to this sampled signal, this N group operator is obtain by cutting successively the filter operator of finite length, the length often organizing operator meet 2 power side and power be more than or equal to 1 positive integer, N >=2;
Merge module 303, for merging the result of calculation of each self-corresponding Branch Computed of this N group operator;
Output module 304, for exporting the result of calculation after merging as the result to this sampled signal.
Optionally, this computing module 302, comprising:
Converter unit 302a, for ought receiving at first and not reached n by the number of the sampled signal of Branch Computed process corresponding to p group operator
ptime, to this n
pindividual sampled signal carries out Fourier transform, obtains n
pindividual Fourier transform data, n
pbe the number of p group operator, 1≤p≤N;
Convolutional calculation unit 302b, for by this n
pindividual Fourier transform data and this p group operator carry out convolutional calculation, obtain n
pindividual convolutional calculation data;
Inverse transform unit 302c, for this n
pindividual convolutional calculation data carry out inverse Fourier transform, obtain n
pindividual inverse Fourier transform data;
Output unit 302d, for by this n
pindividual inverse Fourier transform data export the result of calculation for Branch Computed corresponding to this p group operator.
Optionally, this merging module 303, comprising:
First sum unit 303a, for the n exported by the 1st group of Branch Computed that operator is corresponding
1individual data and the n extracted from First Input First Output corresponding to N group operator
1individual data are sued for peace, by the n that summation obtains
1individual data input the 1st group of First Input First Output that operator is corresponding, the data deficiencies n extracted in the First Input First Output corresponding from N group operator
1time individual, not enough part with 0 polishing, n
1for the number of the 1st group of operator;
Second sum unit 303b, for the n exported by Branch Computed corresponding for this q group operator
qindividual data and the n extracted from First Input First Output corresponding to q-1 group operator
qindividual data are sued for peace, by the n that summation obtains
qindividual data input First Input First Output corresponding to this q group operator, the data deficiencies n extracted in the First Input First Output corresponding from q-1 group operator
qtime individual, not enough part with 0 polishing, n
qbe the number of q group operator, 2≤q≤N;
Wherein, this N group operator is to there being respective First Input First Output.
Optionally, this output module 304, for exporting the data in input the 1st group of First Input First Output that operator is corresponding as this result.
Optionally, this N group operator is the filter coefficient having limit for length's unit impulse response filter.
In sum, the device that the embodiment of the present invention provides, by the operator of convolutional calculation is divided into N group, and carry out convolutional calculation with sampled signal respectively, again the result of convolutional calculation is merged, by one, N number of short deconvolution process synchronously carried out is divided into the long deconvolution process of sampled signal, thus reduces the delay relating to the filtering process of convolutional calculation, improve filter effect.
Please refer to Fig. 6, it illustrates the structural representation of the filter that one embodiment of the invention provides.This filter can be FIR filter.This filter can comprise:
Filter 401 shown in as arbitrary in above-mentioned Fig. 4 or Fig. 5.
In sum, the filter that the embodiment of the present invention provides, by the operator of convolutional calculation is divided into N group, and carry out convolutional calculation with sampled signal respectively, again the result of convolutional calculation is merged, by one, N number of short deconvolution process synchronously carried out is divided into the long deconvolution process of sampled signal, thus reduces the delay relating to the filtering process of convolutional calculation, improve filter effect.
It should be noted that: the filter that above-described embodiment provides is when carrying out filtering process to sampled signal, only be illustrated with the division of above-mentioned each functional module, in practical application, can distribute as required and by above-mentioned functions and be completed by different functional modules, internal structure by device is divided into different functional modules, to complete all or part of function described above.In addition, the filter that above-described embodiment provides and filtering method embodiment belong to same design, and its specific implementation process refers to embodiment of the method, repeats no more here.
The invention described above embodiment sequence number, just to describing, does not represent the quality of embodiment.
One of ordinary skill in the art will appreciate that all or part of step realizing above-described embodiment can have been come by hardware, the hardware that also can carry out instruction relevant by program completes, described program can be stored in a kind of computer-readable recording medium, the above-mentioned storage medium mentioned can be read-only memory, disk or CD etc.
The foregoing is only preferred embodiment of the present invention, not in order to limit the present invention, within the spirit and principles in the present invention all, any amendment done, equivalent replacement, improvement etc., all should be included within protection scope of the present invention.
Claims (10)
1. a filtering method, is characterized in that, described method comprises:
Receive pending sampled signal;
By the N group operator each self-corresponding Branch Computed pre-set, convolutional calculation is carried out to described sampled signal, described N group operator is obtain by cutting successively the filter operator of finite length, the length often organizing operator meet 2 power side and power be more than or equal to 1 positive integer, N >=2;
The result of calculation of each self-corresponding Branch Computed of described N group operator is merged;
Result of calculation after merging is exported as the result to described sampled signal.
2. method according to claim 1, is characterized in that, the described each self-corresponding Branch Computed of N group operator by pre-setting carries out convolutional calculation to described sampled signal, comprising:
When to receive at first and not reached n by the number of the sampled signal of Branch Computed process corresponding to p group operator
ptime, to described n
pindividual sampled signal carries out Fourier transform, obtains n
pindividual Fourier transform data, n
pfor the number of described p group operator, 1≤p≤N;
By described n
pindividual Fourier transform data and described p group operator carry out convolutional calculation, obtain n
pindividual convolutional calculation data;
To described n
pindividual convolutional calculation data carry out inverse Fourier transform, obtain n
pindividual inverse Fourier transform data;
By described n
pindividual inverse Fourier transform data export the result of calculation for Branch Computed corresponding to described p group operator.
3. method according to claim 2, is characterized in that, described N group operator is to there being respective First Input First Output; The described result of calculation to each self-corresponding Branch Computed of described N group operator merges, and comprising:
For the 1st group of operator, by the n that the described 1st group of Branch Computed that operator is corresponding exports
1individual data and the n extracted from First Input First Output corresponding to N group operator
1individual data are added, and will be added the n obtained
1the described 1st group of First Input First Output that operator is corresponding of individual data input, the data deficiencies n extracted in the First Input First Output corresponding from described N group operator
1time individual, not enough part with 0 polishing, n
1for the number of described 1st group of operator;
For q group operator, by the n that Branch Computed corresponding for described q group operator exports
qindividual data and the n extracted from First Input First Output corresponding to q-1 group operator
qindividual data are added, and will be added the n obtained
qthe First Input First Output that the described q group operator of individual data input is corresponding, the data deficiencies n extracted in the First Input First Output corresponding from described q-1 group operator
qtime individual, not enough part with 0 polishing, n
qfor the number of described q group operator, 2≤q≤N.
4. method according to claim 3, is characterized in that, described is the result to described sampled signal by the result of calculation output after merging, comprising:
Data in the described 1st group of First Input First Output that operator is corresponding of input are exported as described result.
5., according to the arbitrary described method of Claims 1-4, it is characterized in that, the filter operator of described finite length is the filter coefficient having limit for length's unit impulse response filter.
6. a filter, is characterized in that, described device comprises:
Receiver module, for receiving pending sampled signal;
Computing module, for the N group operator each self-corresponding Branch Computed by pre-setting, convolutional calculation is carried out to described sampled signal, described N group operator is obtain by cutting successively the filter operator of finite length, the length often organizing operator meet 2 power side and power be more than or equal to 1 positive integer, N >=2;
Merge module, for merging the result of calculation of each self-corresponding Branch Computed of described N group operator;
Output module, for exporting the result of calculation after merging as the result to described sampled signal.
7. device according to claim 6, is characterized in that, described computing module, comprising:
Converter unit, for ought receiving at first and not reached n by the number of the sampled signal of Branch Computed process corresponding to p group operator
ptime, to described n
pindividual sampled signal carries out Fourier transform, obtains n
pindividual Fourier transform data, n
pfor the number of described p group operator, 1≤p≤N;
Convolutional calculation unit, for by described n
pindividual Fourier transform data and described p group operator carry out convolutional calculation, obtain n
pindividual convolutional calculation data;
Inverse transform unit, for described n
pindividual convolutional calculation data carry out inverse Fourier transform, obtain n
pindividual inverse Fourier transform data;
Output unit, for by described n
pindividual inverse Fourier transform data export the result of calculation for Branch Computed corresponding to described p group operator.
8. device according to claim 7, is characterized in that, described merging module, comprising:
First sum unit, for the n exported by the 1st group of Branch Computed that operator is corresponding
1individual data and the n extracted from First Input First Output corresponding to N group operator
1individual data are sued for peace, by the n that summation obtains
1the described 1st group of First Input First Output that operator is corresponding of individual data input, the data deficiencies n extracted in the First Input First Output corresponding from described N group operator
1time individual, not enough part with 0 polishing, n
1for the number of described 1st group of operator;
Second sum unit, for the n exported by Branch Computed corresponding for q group operator
qindividual data and the n extracted from First Input First Output corresponding to q-1 group operator
qindividual data are sued for peace, by the n that summation obtains
qthe First Input First Output that the described q group operator of individual data input is corresponding, the data deficiencies n extracted in the First Input First Output corresponding from described q-1 group operator
qtime individual, not enough part with 0 polishing, n
qfor the number of described q group operator, 2≤q≤N;
Wherein, described N group operator is to there being respective First Input First Output.
9. device according to claim 8, is characterized in that,
Described output module, for exporting the data in the described 1st group of First Input First Output that operator is corresponding of input as described result.
10., according to the arbitrary described device of claim 6 to 9, it is characterized in that, the filter operator of described finite length is the filter coefficient having limit for length's unit impulse response filter.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410584151.6A CN104363003A (en) | 2014-10-27 | 2014-10-27 | Filtering method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410584151.6A CN104363003A (en) | 2014-10-27 | 2014-10-27 | Filtering method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN104363003A true CN104363003A (en) | 2015-02-18 |
Family
ID=52530235
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410584151.6A Pending CN104363003A (en) | 2014-10-27 | 2014-10-27 | Filtering method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104363003A (en) |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2014047854A1 (en) * | 2012-09-27 | 2014-04-03 | 华为技术有限公司 | Adaptive filtering method and system based on error sub-band |
| CN103716013A (en) * | 2014-01-14 | 2014-04-09 | 苏州大学 | Variable parameter proportion self-adaptive filter |
-
2014
- 2014-10-27 CN CN201410584151.6A patent/CN104363003A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2014047854A1 (en) * | 2012-09-27 | 2014-04-03 | 华为技术有限公司 | Adaptive filtering method and system based on error sub-band |
| CN103716013A (en) * | 2014-01-14 | 2014-04-09 | 苏州大学 | Variable parameter proportion self-adaptive filter |
Non-Patent Citations (1)
| Title |
|---|
| VAGNER S.ROSA等: "A High Performance Parallel FIR Filters Generation Tool", 《PROCEEDINGS OF THE SEVENTEENTH IEEE INTERNATIONAL WORKSHOP ON RAPID SYSTEM PROTOTYPING》 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102386889B (en) | Baseline shift minimizing technology, device and median filter | |
| Liu et al. | Design and FPGA implementation of a reconfigurable digital down converter for wideband applications | |
| CN104467739B (en) | The adjustable digital filter of a kind of bandwidth, center frequency point and its implementation | |
| CN107769755B (en) | A design method of parallel FIR decimation filter and parallel FIR decimation filter | |
| CN105281708B (en) | A kind of high speed FIR filter achieving method based on segmentation parallel processing | |
| CN105445550B (en) | A kind of broadband real-time spectrum analysis system and method based on non-blind area digital channelizing | |
| CN113346871B (en) | Multichannel multiphase multi-rate adaptive FIR digital filtering processing architecture | |
| CN104333348B (en) | High-order digital filtering system and high-order digital filtering method | |
| CN110208755B (en) | Dynamic radar echo digital down conversion system and method based on FPGA | |
| CN110620566A (en) | FIR filtering system based on combination of random calculation and remainder system | |
| CN104363003A (en) | Filtering method and device | |
| CN105403769A (en) | Circuit structure based on FFT short-time Fourier analysis and control method thereof | |
| CN102510272B (en) | Method for realizing frequency spectrum sensing by using multi-phase filter | |
| CN106130581A (en) | A kind of multiphase filtering wideband digital channel receiver improves system | |
| Li | Design and realization of FIR digital filters based on MATLAB | |
| CN104539261A (en) | Arbitrary sampling rate conversion interpolation filtering processing method | |
| CN114337764A (en) | A Universal Method and System for a Digital Channelized Receiver Based on Polyphase DFT | |
| CN105515548B (en) | Multichannel based on FPGA extracts the method and device of multiplex filter | |
| Hasan et al. | Improved parameterized efficient FPGA implementations of parallel 1-D filtering algorithms using Xilinx System Generator | |
| CN104579239B (en) | A kind of filter method of filtering system | |
| Deng | Novel iterative second-order-cone-programming scheme for designing high-accuracy phase-circuits | |
| KR101297085B1 (en) | Apparatus of variable fast furier transform and method thereof | |
| US10312954B1 (en) | Identification of RFI (radio frequency interference) | |
| CN111262598B (en) | Complex baseband signal reconstruction processing method and device | |
| Madheswaran et al. | Implementation And Comparison Of Different CIC Filter Structure For Decimation |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C41 | Transfer of patent application or patent right or utility model | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20170221 Address after: 510000 B1, building, No. 16, rhyme Road, Guangzhou, Guangdong, China 13F Applicant after: Guangzhou KuGou Networks Co., Ltd. Address before: 213161 Jiangsu Province, Changzhou city Wujin district north of the city and Zhongzhi 328 floor Applicant before: CHANGZHOU HEARING WORKSHOP INTELLIGENT TECHNOLOGY CO., LTD. |
|
| CB02 | Change of applicant information |
Address after: 510660 Guangzhou City, Guangzhou, Guangdong, Whampoa Avenue, No. 315, self - made 1-17 Applicant after: Guangzhou KuGou Networks Co., Ltd. Address before: 510000 B1, building, No. 16, rhyme Road, Guangzhou, Guangdong, China 13F Applicant before: Guangzhou KuGou Networks Co., Ltd. |
|
| CB02 | Change of applicant information | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20150218 |
|
| RJ01 | Rejection of invention patent application after publication |