Method, system and computer-readable storage medium for integer signal spectrum sampling and restoration
By sampling and transmitting a subset of the integer signal spectrum using FFT, the method enhances transmission efficiency and reduces hardware requirements, achieving lossless compression with improved performance.
Patent Information
- Application Number
- JP2023069245
- Authority / Receiving Office
- JP · JP
- Patent Type
- Patents
- Current Assignee / Owner
- Priority Date
- 2023-02-24
- Filing Date
- 2023-04-20
- Publication Date
- 2024-11-12
- Estimated Expiration
- 2043-04-20
AI Technical Summary
Existing lossless compression techniques require sampling about 33% of the original signal length, leaving room for improvement in performance and efficiency.
The method involves performing fast Fourier transform on an integer signal to obtain its spectrum, sampling a subset of this spectrum, and transmitting the sampled samples to reconstruct the original integer signal, utilizing a computer-readable storage medium and processors for this process.
This approach allows for transmitting only a part of the integer signal spectrum, saving transmission frequency band and improving efficiency by reducing the required data amount, while achieving lossless compression without needing dictionary data.
Smart Images

Figure 0007581410000001 
Figure 0007581410000002 
Figure 0007581410000003
Abstract
Description
Technical Field
[0001] The present invention relates to a reversible compression technique applicable to signal transmission, and particularly to a technique for spectral sampling and restoration of integer signals.
Background Art
[0002] In conventional transmission compression methods, it is common to remove information related to a certain frequency band by taking advantage of the fact that the human eye and ear are not sensitive to it. Even if the restored signal is different from the original signal, it will not be detected. Such compression is called irreversible compression.
[0003] On the other hand, the compression may be dictionary-based compression. For example, an N×M matrix (N>>M, that is, N is much larger than M) is constructed in advance, and using the concept of compressed sensing, an M-point signal is converted into N dimensions. According to this method, although the dimension seems to be expanded, only K-point values (K<<M, that is, K is much smaller than M) exist in this N dimension, so the compression effect is achieved and it belongs to reversible compression.
Summary of the Invention
Problems to be Solved by the Invention
[0004] There is also a technique different from the above two methods. First, this method is a reversible compression in which the restored signal exactly matches the original signal. Second, since this technique does not require the construction of dictionary data, the storage capacities of the transmitting device and the receiving device are saved. However, since such a technique needs to sample about 33% of the original signal length, there is still room for improvement in its performance.
Means for Solving the Problems
[0005] To solve the above problems, the present invention provides an original integerThe present invention provides a method for sampling and restoring a spectrum of an integer signal, the method including the steps of: acquiring a spectrum of an original integer signal, which is a sequence of N (N=a power of 2) integers equal to or greater than 0, by performing a fast Fourier transform (FFT) on the signal; sampling the spectrum and assembling a plurality of sampled samples into an original sample sequence; transmitting the original sample sequence and restoring the original integer signal based on the original sample sequence.
[0006] The present invention also provides a computer readable storage medium having stored thereon instructions for being loaded into a computing or electronic device to cause the device to carry out the above-described integer signal spectrum sampling and restoration method.
[0007] The present invention further provides a system for integer signal spectrum sampling and restoration, comprising a sending device including a first processor and a transmitting module, and a receiving device including a second processor and a receiving module, wherein the first processor of the sending device performs a fast Fourier transform on an original integer signal, which is a sequence of N (N=a power of 2) integers equal to or greater than 0, to obtain a spectrum of the original integer signal, the first processor of the sending device further samples the spectrum and assembles the sampled samples into an original sample sequence, the transmitting module of the sending device transmits the original sample sequence, the receiving module of the receiving device receives the original sample sequence, the second processor of the receiving device restores the original integer signal based on the original sample sequence, and the receiving device stores the original sample sequence and the original integer signal. Effect of the Invention
[0008] As described above, the present invention can achieve the effect of saving transmission frequency bands, since when transmitting an integer signal, only a part of the integer signal spectrum needs to be transmitted instead of the entire integer signal spectrum. [Brief description of the drawings]
[0009] [Figure 1] FIG. 1 is a schematic flow diagram illustrating a method for integer signal spectrum sampling and restoration according to an embodiment of the present invention. [Diagram 2] FIG. 2 is a schematic flow diagram of a method for restoring an integer signal according to an embodiment of the present invention. [Diagram 3] FIG. 3 is a schematic block diagram of a transmitting device and a receiving device according to an embodiment of the present invention. [Figure 4] FIG. 4 is a schematic block diagram of a transmitting device and a receiving device according to an embodiment of the present invention. [Diagram 5] FIG. 5 is a schematic block diagram of a transmitting device and a receiving device according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0010] The following describes the implementation of the present invention through specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents described herein.
[0011] 1 is a schematic flow chart of an integer signal spectrum sampling and restoration method according to an embodiment of the present invention, which can be divided into an integer signal spectrum sampling method and an integer signal restoration method.
[0012] First, the sending device 100 executes the integer signal spectrum sampling method of steps 110 to 140, converts the input integer signal into a discrete Fourier transform (DFT), performs sampling on the DFT to obtain a sample sequence, and sends the sample sequence to the receiving device 200.
[0013] Then, the receiving device 200 executes the integer signal restoration method in steps 210 to 220 to receive the sample sequence and restore the sample sequence to an original integer signal.
[0014] In one embodiment, the sending device 100 and the receiving device 200 may be computing devices or electronic devices having data processing, calculation, transmission and reception functions. The sending device 100 and the receiving device 200 are communicatively or electrically connected to each other. The flow of FIG. 1 will be described below.
[0015] First, in step 110, an integer signal is input to the transmitting device 100. The integer signal is an N-point integer sequence consisting of N (N=power of 2, e.g., 32, 64, 128, or 256) sequences of integers x(0), x(1), . . . , x(N-1) that are equal to or greater than 0.
[0016] Next, in step 120, a discrete Fourier transform of the integer signal x(0), x(1), ..., x(N-1) is obtained by fast Fourier transform. In general, x(0), x(1), ..., x(N-1) are all complex numbers. The discrete Fourier transform of the integer signal is also called the spectrum of the integer signal.
[0017] Then, in step 130, a sampling method is performed on the spectrum to obtain a sample sequence of the spectrum. In detail, if the value of N is small, for example, equal to or smaller than M (M is also a power of 2), the complex numbers whose serial numbers are 0 in the spectrum, and all complex numbers whose serial numbers are greater than 0 and smaller than N and can be divisible by N, can be selected as samples in the sample sequence.
[0018] In one embodiment, N=32, the complex numbers with serial numbers 0, 1, 2, 4, 8, and 16 in the spectrum are selected as samples, i.e., the sample sequence consists of six complex numbers in the spectrum: x(0), x(1), x(2), x(4), x(8), and x(16). In the sample sequence, these samples are ordered in ascending order of serial numbers in the spectrum.
[0019] Hereinafter, steps 110, 120, and 130 will be described using an actual 32-point two-dimensional signal as an example. 1,0,0,1,0,0,1,0,1,1,0,1,0,1,1,1, 1,0,0,1,0,0,0,0,1,1,0,0,0,0,0,0 Suppose you input:
[0020] Spectrum obtained by FFT X(k) = 13.0000 +0i, -2.9090 - 2.8887i, -0.1585 + 1.1481i, -0.9622 - 2.2726i, 3.2929 - 0.1213i, 1.5367 - 0.5870i, -2.2304 + 2.7716i, 0.0702 - 2.0114i, 2.0000 + 1.0000i, 1.0122 + 0.6017i, 2.2304 + 2.7716i, 1.0764 + 0.4954i, 4.7071 - 4.1213i, -1.6510 - 1.1902i, 0.1585 + 1.1481i, 1.8266 - 0.2756i, -1.0000 + 0i, 1.8266 + 0.2756i, 0.1585 - 1.1481i, -1.6510 + 1.1902i, 4.7071 + 4.1213i, 1.0764 - 0.4954i, 2.2304 - 2.7716i, 1.0122 - 0.6017i, 2.0000 - 1.0000i, 0.0702 + 2.0114i, -2.2304 - 2.7716i, 1.5367 + 0.5870i, 3.2929 + 0.1213i, -0.9622 + 2.2726i, -0.1585 - 1.1481i, -2.9090 + 2.8887i It is.
[0021] Normally, all X(k) (k=0 to 31) are required to restore the integer signal x(n). However, according to the method of the present invention, the integer signal x(n) can be restored simply by extracting X(k) with k=0, 1, 2, 4, 8, and 16, i.e., X(0)=13, X(1)=-2.9090-2.8887i, X(2)=-0.1585 +1.1481i, X(4)=3.2929-0.1213i, X(8)=2.0000+1.0000i, and X(16)=-1.0000+0i.
[0022] On the other hand, if the value of N is large, for example greater than M, then 0, all integers greater than 0 and less than M that are divisible by M, and integers obtained by adding M×j (j is an integer between 1 and d−1, and d=N / M) to each of the above integers (including 0) can be selected as serial numbers in the spectrum of a sample in the sample sequence.
[0023] In another embodiment, let N=128 and M=64, then 0, 1, 2, 4, 8, 16, 32, 64, 65, 66, 68, 72, 80, 96 are selected as sample serial numbers, where 1, 2, 4, 8, 16, 32 are all integers greater than 0 and less than M that can be divided into M. The serial numbers 64, 65, 66, 68, 72, 80, 96 are integers obtained by adding 64 to the serial numbers 0, 1, 2, 4, 8, 16, 32, respectively. Therefore, the sample sequence consists of 14 complex numbers in the spectrum, namely X(0), X(1), X(2), X(4), X(8), X(16), X(32), X(64), X(65), X(66), X(68), X(72), X(80), X(96).
[0024] In yet another embodiment, where N=256 and M=64, in addition to the 14 sample sequences with N=128, serial numbers 0, 1, 2, 4, 8, 16, and 32 are added with 128 (64×2) and 192 (64×3), respectively, to generate another 14 serial numbers, totaling 28 serial numbers, which are composed of 28 complex numbers in the spectrum corresponding to the 28 serial numbers.
[0025] In the above two embodiments, the data amount of the sample sequence obtained by sampling is much smaller than the complete spectrum before sampling, so that the effect of data compression can be achieved and the overall performance can be improved.
[0026] Next, in step 140, the sending device 100 transmits the sample sequence to the receiving device 200 via wired or wireless communication.
[0027] In step 210, the receiving device 200 receives the sample sequence.
[0028] Then, in step 220 , the receiving device 200 executes a restoration method to obtain an original integer signal based on the sample sequence, and inputs the integer signal of the sending device 100 in step 110 .
[0029] The restoration method in step 220 is shown in FIG.
[0030] First, in step 221, the sample sequence is input to the reconstruction method.
[0031] Next, in step 222, it is determined whether the length N of the original integer signal is greater than M; if N is less than or equal to M, proceed to step 224; if not, proceed to step 223.
[0032] In step 224, a short point number restoration method is performed to restore the sample sequence to the original integer signals x(0), x(1), ..., x(N-1) to achieve the purpose of data transmission. In one embodiment, the short point number restoration method in step 224 can use brute force attack, branch and bound algorithm or any other techniques that can solve mixed integer linear programming problems, and the present invention is not limited thereto.
[0033] In one embodiment, for N=32, after the receiving device 200 receives the sample sequence of six complex numbers, X(0), X(1), X(2), X(4), X(8), and X(16), in step 221, since N=32 is less than M=64, the flow proceeds to step 224, where the short point number reconstruction method is performed to reconstruct integer signals x(0), x(1), ..., x(31) based on X(0), X(1), X(2), X(4), X(8), and X(16).
[0034] Meanwhile, in step 223, if the length of the sample sequence is L (i.e., it contains L complex samples), generate d sample sequences with length L / d based on the sample sequence, and divide the original N-point mixed integer problem into d M-point mixed integer problems, where d=N / M. Then, in step 224, use a short point number reconstruction method to reconstruct d M-point integer signals based on each of the d sample sequences with length L / d, and assemble the d M-point integer signals into the original N-point integer signal.
[0035] In another embodiment, for N=128, after receiving a sample sequence of L=14 complex numbers in step 221, since N=128 is greater than M=64 and d=128 / 64=2, the 128-point problem is first divided into two 64-point problems in step 223. Then, in step 224, the two 64-point integer signals are restored by a short point number restoration method, and the two 64-point integer signals are reassembled into an original 128-point integer signal.
[0036] As above, if d=2, the details of the division in step 223 are as follows:
[0037] First, two sample sequences Ev(k) and 0d(k) each having a length of L / 2 can be obtained based on an original sample sequence having a length of L and the following formula: Ev(k)=[X(k)+X(k+M)] / 2 Od(k)=[X(k)-X(k+M)] / 2 / W
[0038] Here, k is 0 and / or any integer greater than 0 and less than M that divides M. For M=64, k=0, 1, 2, 4, 8, 16, 32. Also, W=exp(-2πi / N).
[0039] According to the properties of the Fourier transform, Ev(k), k = 0, 1, 2, 4, 8, 16, 32 corresponds to the sample sequence obtained by taking a 64-point Fourier transform of the 64-point sequence of even serial number points x(2n), n = 0, 1, 2, ..., 63 in the original integer signal, and then sampling at serial numbers k = 0, 1, 2, 4, 8, 16, 32. In other words, this is a 64-point mixed integer problem.
[0040] Similarly, Od(k), k=0,1,2,4,8,16,32 corresponds to the sample sequence obtained by taking a 64-point Fourier transform of the 64-point sequence of odd serial number points x(2n+1), n=0,1,2,...,63 in the original integer signal, and then sampling at serial numbers k=0,1,2,4,8,16,32. This is another independent 64-point mixed integer problem.
[0041] After recovering two 64-point integer signals according to Ev(k) and Od(k), the original 128-point integer signal can be obtained by cross-arranging the integers in the two integer signals. Assuming that M is another power of 2, an analogy can be made on this basis.
[0042] The above embodiment and flow convert one 128-point problem into two 64-point problems. When a brute force attack is used in the restoration of step 224, two 64-point problems are converted into 128-point problems. 128 It is necessary to use a combination of seeds. For two divided 64-point problems, 65 =2×2 64 By simply using the seed combination, brute force attacks can be performed, which greatly reduces the difficulty of recovery and improves overall performance.
[0043] Also, when d is 2 or other numerical value, the division of the mixed integer problem in step 223 can be processed based on general knowledge. For example, see the paper "On Hardware Implementation of Radix 3 and Radix 5 FFT Kernels for LTE Systems" by Johan Lofgren and Peter Nilsson, published in the IEEE periodical "2011 NORCHIP" in November 2011. The present invention is not limited thereto.
[0044] 3 to 5 are schematic block diagrams of a sending device 100 and a receiving device 200 according to an embodiment of the present invention.
[0045] 3, the sending device 100 includes a storage device 101, a processor 102, and a sending module 103, where the processor 102 is communicatively connected to the storage device 101 and the sending module 103. The receiving device 200 includes a storage device 201, a processor 202, and a receiving module 203, where the processor 202 is communicatively connected to the storage device 201 and the receiving module 203.
[0046] In one embodiment, the storage device 101 and the storage device 201 may both be memory, a hard disk or other hardware device capable of storing data or information, and the sending module 103 and the receiving module 203 may both be modules consisting of hardware, software or firmware.
[0047] 4, the processor 102 of the sending device 100 includes a conversion module 151 and a sampling module 152 that are communicatively connected to each other. The conversion module 151 and the sampling module 152 may both be modules made of hardware, software, or firmware.
[0048] 5, the processor 202 of the receiving device 200 includes an input module 251, a division module 252, and a restoration module 253, which are communicatively connected to each other. The input module 251, the division module 252, and the restoration module 253 may each be a module made of hardware, software, or firmware.
[0049] The sending device 100 and the receiving device 200 according to this embodiment are used to implement the integer signal spectrum sampling method and the integer signal restoration method described above.
[0050] Specifically, as shown in Figures 1 and 3, the storage device 101 of the sending device 100 can store the integer signal of the sending device 100 in step 110. The processor 102 can be used to perform steps 120 and 130, where the conversion module 151 performs step 120, and the sampling module 152 performs step 130. In steps 140 and 210, the sending module 103 of the sending device 100 can send the sample sequence to the receiving module 203 of the receiving device 200 by wire or wirelessly.
[0051] Also, the storage device 201 of the receiving device 200 can store the sample sequence received by the receiving module 203 in step 210. The processor 202 can be used to perform step 220. Here, the input module 251 can be used to receive the sample sequence in step 221 and perform the decision in step 222. Also, the division module 252 can be used to perform step 223, and the restoration module 253 can be used to perform step 224 and store the restored original integer signal in the storage device 201.
[0052] In another embodiment, the present invention provides a computer readable storage medium, including at least one memory, floppy disk, hard disk and / or magneto-optical disk, storing instructions that can be read into the sending device 100 and the receiving device 200 or other computing or electronic device to perform the above integer signal spectrum sampling and restoration method.
[0053] The present invention also provides a lossless compression technique applicable to signal transmission. In the process of integer signal spectrum sampling and restoration, the number of sampled samples varies according to the length of the integer signal sequence, and different measures are implemented in the restoration according to the length of the integer signal sequence.
[0054] In addition, if the integer signal sequence is long, it is decomposed into several short sequence signals and restored individually, and if the integer signal sequence is short, the original integer signal is obtained directly.
[0055] As described above, the present invention transmits only a part of the integer signal spectrum instead of the entire integer signal spectrum, compressing the integer signal to save the transmission frequency band between the transmitting device and the receiving device and improve the transmission efficiency between the transmitting device and the receiving device, and the receiving device can restore the integer signal without distortion by the restoration method of the present invention.
[0056] Thus, the present invention can achieve the restoration of an integer signal sequence consisting of 64 0 / 1 bit values within 0.03 seconds using a sampling rate of about 11%, which is a great improvement over the conventional lossless compression technology that requires a sampling rate of 33%, and not only can it significantly reduce the hardware resources required for compression and restoration, such as processors and storage devices (similar to the above modules), but it also has the advantages of lossless compression and the advantage of not requiring dictionary data, thereby improving the overall performance.
[0057] The above-mentioned embodiments are merely illustrative of the principles and effects of the present invention, and are not intended to limit the present invention, and those skilled in the art may make various modifications and changes to the embodiments without departing from the spirit of the present invention. Therefore, the scope of protection of the rights of the present invention is included in the scope of the following claims. [Explanation of symbols]
[0058] 100 Sending device 101 Storage device 102 processors 103 Transmitting Module 110, 120, 130, 140 steps 151 Conversion Module 152 Sampling Module 200 Receiving device 201 Storage device 202 Processor 203 Receiver Module 210, 220~224 steps 251 Input Module 252 division module 253 Recovery Module
Claims
1. performing a fast Fourier transform on the original integer signal to obtain a spectrum of the original integer signal, which is a sequence of N integers (N=a power of 2) equal to or greater than 0; sampling the spectrum and assembling the sampled samples into an original sample sequence; transmitting the original sample sequence and recovering the original integer signal based on the original sample sequence; Including, The serial numbers in the spectrum of the samples include d (d=N / M) sets of serial numbers, where M is a power of 2 less than N, a 0th set of serial numbers in the d sets of serial numbers includes 0 and / or all integers greater than 0 and less than M that are divisible by M, and a jth set of serial numbers in the d sets of serial numbers includes integers obtained by adding M×j (j is any integer from 1 to d−1) to each serial number in the 0th set of serial numbers.
2. 2. The method of claim 1, wherein in the original sample sequence, the samples are ordered based on their serial numbers in the spectrum.
3. 2. The method of claim 1, wherein the serial numbers in the spectrum of those samples include 0 and / or all integers greater than 0 and less than N and divisible by N.
4. 4. The method of claim 3, further comprising performing a short point number restoration method on the original sample sequence to restore the original integer signal.
5. generating, based on the original sample sequence, a sequence of d short point number samples, the number of samples contained being d times smaller than the number of samples contained in the original sample sequence; reconstructing d short point integer signals, each of which is a sequence of M integers in the original integer signal, based on the short point sample sequences using a short point reconstruction method; assembling the short point integer signals into the original integer signal; 2. The method of claim 1, further comprising:
6. The method for integer signal spectrum sampling and restoration according to claim 5, characterized in that, when d=2, each sample in each short point number sample sequence is calculated based on two samples in the original sample sequence whose serial numbers differ by M.
7. 6. The method of claim 5, further comprising the step of reconstructing the original integer signal by cross-arranging the integers in the short point integer signal.
8. A computer-readable storage medium having stored thereon instructions that, when loaded into a computing or electronic device, cause the method of integer signal spectrum sampling and restoration according to any one of claims 1 to 7 to be carried out.
9. a transmitting device including a first processor that performs a fast Fourier transform on an original integer signal to obtain a spectrum of the original integer signal, the spectrum being a sequence of N (N is a power of 2) integers equal to or greater than 0, and that performs sampling on the spectrum to assemble the sampled samples into an original sample sequence; and a transmitting module that transmits the original sample sequence; a receiving device communicatively or electrically connected to the transmitting device, the receiving device including a receiving module for receiving the original sample sequence and a second processor for reconstructing the original integer signal based on the original sample sequence, the receiving device storing the original sample sequence and the original integer signal; Equipped with The serial numbers in the spectrum of the samples include d (d=N / M) sets of serial numbers, where M is a power of 2 less than N, a 0th set of serial numbers in the d sets of serial numbers includes 0 and / or all integers greater than 0 and less than M that are divisible by M, and a jth set of serial numbers in the d sets of serial numbers includes integers obtained by adding M×j (j is any integer from 1 to d−1) to each serial number in the 0th set of serial numbers.
Citation Information
Patent Citations
Image data transmission and reproduction system, image data transmission and reproduction method, program, and storage medium
JP2004260524A
Memory Access Windows and Padding for Motion Vector Refinement and Motion Compensation
JP2023014095A
Image resampling for DCT based image encoding formats using memory efficient techniques
US20210067782A1
