A calculation method and system for optoelectronic hybrid tensor convolution

Abstract

The present invention discloses a calculation method and system for optoelectronic hybrid tensor convolution, relating to the technical field of optical computing. The method includes the following steps: splitting the acquired original optical carrier and adjusting it into two optical carriers with the same frequency and in-phase; loading the input information onto the two optical carriers and adjusting them into two optical signals with equal optical paths; mixing the two optical signals to generate two optical mixing signals with phase differences of 0° and 180° respectively; performing optoelectronic conversion on the two optical mixing signals and determining the convolution result according to the output electrical signals. The present invention can obtain a complete convolution result in one operation, the amplitudes and frequencies of the input signals are adjustable, and the relevant instruments and equipment can be controlled by a computer, meeting the requirements of large throughput and flexibility.

Description

Technical Field

[0001] The present invention relates to the field of optical computing technology, and particularly relates to a method and system for optoelectronic hybrid tensor convolution calculation. Background Art

[0002] An artificial neural network is a collection of nodes with weighted connections. These nodes can modify the network weights through appropriate feedback, and can "learn" and perform complex operations for face recognition, speech translation, medical diagnosis, etc. However, whether it is a traditional fully connected feedforward network or currently widely concerned convolutional neural networks, recurrent neural networks, etc., a large number of tensor convolution operations need to be performed. These operations account for 55% to 90% of the total computing volume. Therefore, accelerating tensor convolution operations is an important way to improve the computing power of neural networks.

[0003] To improve the computing power of neural networks, accelerating tensor convolution operations must improve the computing power of the underlying neuromorphic hardware. The Optical Neural Network (ONN) has the potential to overcome the bandwidth bottleneck of electronic neural networks, and has the potential of high speed and low power consumption. It is one of the development trends of the next generation of neuromorphic computing.

[0004] Currently, there are mainly three methods to implement optical tensor convolution operations: the first is to load the data of the weight tensor onto the amplitude of the optical frequency comb, then use an electro-optic modulator to modulate the data of the input tensor onto the optical frequency comb, and use the optical fiber to introduce dispersion to complete the tensor convolution calculation; the second is to map the data of the weight tensor onto the phase change material array, and load the data of the input tensor onto the amplitude of the optical frequency comb teeth. After the optical frequency comb passes through the phase change material array, the tensor convolution calculation is completed; the third is to use an array of Mach-Zehnder interference units to represent the weight tensor, load the data of the input tensor onto a coherent or incoherent light source, and complete the convolution operation after passing through this array. Among them, the coherent light source can complete complex domain operations, and the incoherent light source realizes real domain operations.

[0005] However, due to the requirements of its own implementation, it is difficult to increase the tensor size and the number of tensors in the first method above. Moreover, due to the need to utilize the optical fiber dispersion effect, it requires a large physical space, and the effective computing efficiency needs to be improved; in the second method, due to being restricted by the scale of the phase change material array and the number of available optical frequency comb teeth, the tensor size and the number of tensor cores need to be increased; in the third method, as the tensor scale increases, it will be restricted by the integration scale and it is difficult to meet the need to implement large-scale tensor convolution operations. Summary of the Invention

[0006] Aiming at the defects existing in the prior art, the first aspect of the present invention provides a calculation method for optoelectronic hybrid tensor convolution, which can obtain a complete convolution result in one operation and the amplitude and frequency of the input electrical signal are adjustable, meeting the requirements of large throughput and flexibility.

[0007] To achieve the above object, the technical solution adopted by the present invention is:

[0008] A calculation method for optoelectronic hybrid tensor convolution, the method comprising the following steps:

[0009] Split the obtained original optical carrier and adjust it into two optical carriers with the same frequency and phase;

[0010] Load the input information onto the two optical carriers and adjust them into two optical signals with equal optical paths;

[0011] Mix the two optical signals to generate two optical mixing signals with phase differences of 0° and 180° respectively;

[0012] Perform optoelectronic conversion on the two optical mixing signals and determine the convolution result according to the output electrical signal.

[0013] In some embodiments, the splitting the obtained original optical carrier and adjusting it into two optical carriers with the same frequency and phase includes:

[0014] Obtain the original optical carrier, and the original optical carrier is a single-wavelength narrow-linewidth laser carrier;

[0015] Split the original optical carrier and adjust the optical path to obtain two optical carriers with the same frequency and phase.

[0016] In some embodiments, the loading the input information onto the two optical carriers and adjusting them into two optical signals with the same phase includes:

[0017] Generate two corresponding analog electrical signals according to the input data;

[0018] Generate two electrical drive signals with a phase difference of 90° based on the two analog electrical signals;

[0019] Load the input data onto the two optical carriers according to the electrical drive signals and the corresponding optical carriers;

[0020] Adjust the optical path of the modulated optical carrier to generate two optical signals with equal optical paths.

[0021] The second aspect of the present invention lies in providing an optoelectronic hybrid tensor convolution calculation system, which can obtain a complete convolution result in one operation and the amplitude and frequency of the input electrical signal are adjustable, meeting the requirements of large throughput and flexibility.

[0022] To achieve the above object, the technical solution adopted by the present invention is:

[0023] An optoelectronic hybrid tensor convolution calculation system, comprising:

[0024] A laser source for generating an original optical carrier;

[0025] A first optical delay and beam splitter device for splitting the original optical carrier and adjusting it into two optical carriers with the same frequency and phase;

[0026] An electro-optic modulation device for loading input information onto the two optical carriers;

[0027] A second optical delay device for adjusting the two modulated optical carriers into two optical signals with equal optical paths;

[0028] An optical mixer for mixing the two optical signals to generate two optical mixing signals with phase differences of 0° and 180° respectively;

[0029] An optoelectronic balanced detector for performing optoelectronic conversion on the two optical mixing signals to output an electrical signal;

[0030] A data acquisition device for acquiring the output electrical signal and determining the convolution result based on the acquired electrical signal.

[0031] In some embodiments, the first optical delay and beam splitter device includes:

[0032] A beam splitter for splitting the original optical carrier into two optical carriers;

[0033] A first tunable optical delay line, whose input end is connected to the beam splitter, for adjusting the optical path of the optical carrier.

[0034] In some embodiments, the first optical delay and beam splitter device further includes:

[0035] A first fiber optic amplifier, the input end of the first fiber optic amplifier is connected to the output end of the laser source, and the output end of the first fiber optic amplifier is connected to the input end of the beam splitter.

[0036] In some embodiments, the electro-optic modulation device includes:

[0037] A first modulation component, which includes a first waveform generator, a first modulator driver, a first 90° electrical mixer, and a first IQ modulator connected in sequence, and the input end of the first IQ modulator is directly connected to the output end of the beam splitter;

[0038] A second modulation component, which includes a second waveform generator, a second modulator driver, a second 90° electrical mixer, and a second IQ modulator connected in sequence, and an input end of the second IQ modulator is connected to an output end of the first adjustable optical delay line.

[0039] In some embodiments, the second optical delay device includes:

[0040] A second optical fiber amplifier, an input end of the second optical fiber amplifier is connected to an output end of the first IQ modulator, and an output end of the second optical fiber amplifier is connected to an input end of the optical mixer;

[0041] A second adjustable optical delay line, an input end of which is connected to an output end of the second IQ modulator;

[0042] A third optical fiber amplifier, an input end of the third optical fiber amplifier is connected to an output end of the second adjustable optical delay line, and an output end of the third optical fiber amplifier is connected to an input end of the optical mixer.

[0043] In some embodiments, a computer control system is further included, and the computer control system is used to control adjustment parameters of the first adjustable optical delay line, the electro-optical modulation device, and the second adjustable optical delay line.

[0044] In some embodiments, the data acquisition device is used to feed back the convolution result to the computer control system.

[0045] Compared with the prior art, the advantages of the present invention are as follows:

[0046] In the optoelectronic hybrid tensor convolution calculation system of the present invention, the acquired original optical carrier is split and adjusted into two optical carriers with the same frequency and the same phase; and the input information is loaded onto the two optical carriers and adjusted into two optical signals with the same phase. A complete convolution result can be obtained through one operation by the balanced photodetector. Moreover, the adjustment parameters of each adjustable device can be modified through the computer control system, which is convenient to set, simple and convenient. And each optical device adopts optical fiber devices and can be integrated on a chip subsequently. Therefore, it has the advantages of high signal-to-noise ratio, large throughput, and easy subsequent on-chip integration, and well meets the requirements of the optical computing field. Description of the Drawings

[0047] Figure 1 It is a flowchart of the optoelectronic hybrid tensor convolution calculation method in an embodiment of the present invention;

[0048] Figure 2 It is a block diagram of the optoelectronic hybrid tensor convolution calculation system in an embodiment of the present invention;

[0049] Figure 3 It is a schematic structural diagram of the optoelectronic hybrid tensor convolution calculation system in an embodiment of the present invention. Detailed implementation mode

[0050] For the purpose of making the objectives, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Apparently, the described embodiments are only a part rather than all of the embodiments of the present application. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present application without creative efforts shall fall within the scope of protection of the present application.

[0051] See Figure 1 As shown, an embodiment of the present invention provides a calculation method for optoelectronic hybrid tensor convolution, and the method includes the following steps:

[0052] S1. Split the acquired original optical carrier and adjust it into two optical carriers with the same frequency and the same phase.

[0053] Specifically, step S1 includes:

[0054] S11. Acquire the original optical carrier, and the original optical carrier is a single-wavelength narrow-linewidth laser carrier.

[0055] S12. Split the original optical carrier and adjust the optical path to obtain two optical carriers with the same frequency and the same phase.

[0056] It should be noted that in a specific implementation, an optical splitter can be used to split the original optical carrier, and then an adjustable optical delay line can be used to adjust the optical path, so as to obtain two optical carriers with the same frequency and the same phase.

[0057] S2. Load the input information onto the two optical carriers and adjust them into two optical signals with equal optical paths.

[0058] Specifically, step S2 includes:

[0059] S21. Generate two corresponding analog electrical signals according to the input data.

[0060] S22. Generate two electrical driving signals with a phase difference of 90° based on the two analog electrical signals respectively.

[0061] S23. Load the input data onto the two optical carriers according to the electrical driving signals and the corresponding optical carriers.

[0062] S24. Adjust the optical path of the modulated optical carrier to generate two optical signals with equal optical paths.

[0063] It should be noted that in a specific implementation

[0064] S3. Mix the two optical signals to generate two optical mixing signals with phase differences of 0° and 180° respectively.

[0065] S4. Perform photoelectric conversion on the two optical mixing signals, and determine the convolution result according to the output electrical signal.

[0066] Therefore, in the calculation method of optoelectronic hybrid tensor convolution in the present invention, the obtained original optical carrier is split and adjusted into two optical carriers with the same frequency and phase; and the input information is loaded onto the two optical carriers and adjusted into two optical signals with equal optical paths. The complete convolution result can be obtained through one operation of the photodetector. Moreover, the parameters of each adjustable device can be modified through the computer control system, which is convenient to set, simple and convenient.

[0067] See Figure 2 As shown in the figure, an embodiment of the present invention provides a calculation system for optoelectronic hybrid tensor convolution, which includes a laser source 1, a first optical delay and splitting device 2, an electro-optic modulation device 3, a second optical delay device 4, an optical mixer 5, a photoelectric balanced detector 6, a data acquisition device 7, and a computer control system 8.

[0068] Among them, the laser source 1 is used to generate the original optical carrier; the first optical delay and splitting device 2 is used to split and adjust the original optical carrier into two optical carriers with the same frequency and phase; the electro-optic modulation device 3 is used to load the input information onto the two optical carriers; the second optical delay device 4 is used to adjust the two modulated optical carriers into two optical signals with equal optical paths; the optical mixer 5 is used to mix the two optical signals to generate two optical mixing signals with phase differences of 0° and 180° respectively; the photoelectric balanced detector 6 is used to perform photoelectric conversion on the two optical mixing signals to output an electrical signal; the data acquisition device 7 is used to collect the output electrical signal and determine the convolution result according to the collected electrical signal.

[0069] Preferably, the calculation system for optoelectronic hybrid tensor convolution further includes a computer control system, which is used to control the adjustment parameters of the first adjustable optical delay line, the electro-optic modulation device, and the second adjustable optical delay line.

[0070] Furthermore, the data acquisition device 7 is used to feedback the convolution result to the computer control system 8. The computer control system 8 adjusts parameters, displays results, stores data, etc. according to the feedback result.

[0071] Specifically, the laser source 1 provides a single-wavelength narrow-linewidth laser carrier for the system. The first optical delay and beam splitting device 2 amplifies the optical carrier power, splits the light, and adjusts the optical paths of two beams to be consistent for the system, thereby providing two coherent light sources with the same frequency and phase. The electro-optic modulation system 3 based on the IQ modulator modulates the optical carrier according to its input data and loads the input information onto the optical carrier. The second optical delay device 4 is used to adjust the optical paths between the two beams to be consistent and amplify the powers of the two beams. The two input beams are mixed by the 180° optical mixer 5 to generate two signal beams. The output is through the optoelectronic balanced detector 6. The data acquisition device 7 is used to collect the electrical signals and output the results to the computer system 8. The computer control system 8 can control the input analog data of the arbitrary waveform generator, the optical path of the tunable optical delay line, the bias voltage of the IQ modulator, and the detection parameters of the oscilloscope, ultimately realizing the flexible adjustment of the optical computing system.

[0072] See Figure 3 As shown in some embodiments, the first optical delay and beam splitting device 2 includes:

[0073] A beam splitter 202 for splitting the original optical carrier into two optical carriers.

[0074] A first tunable optical delay line 203, whose input end is connected to the beam splitter 202, for adjusting the optical path of the optical carrier. The optical path of the first tunable optical delay line 203 is set by the computer control system 8 to ensure that the two optical signals are in phase.

[0075] Furthermore, the first optical delay and beam splitting device 2 further includes:

[0076] A first fiber optic amplifier 201, whose input end is connected to the output end of the laser source 1, and whose output end is connected to the input end of the beam splitter 202. Preferably, the first fiber optic amplifier 201 is an erbium-doped fiber amplifier.

[0077] In some embodiments, the electro-optic modulation device 3 includes:

[0078] A first modulation component, which includes a first waveform generator 301, a first modulator driver 303, a first 90° electrical mixer 305, and a first IQ modulator 307 connected in sequence, and the input end of the first IQ modulator 307 is directly connected to the output end of the beam splitter 202;

[0079] A second modulation component, which includes a second waveform generator 302, a second modulator driver 304, a second 90° electrical mixer 306, and a second IQ modulator 308 connected in sequence, and the input end of the second IQ modulator 308 is connected to the output end of the first tunable optical delay line 203.

[0080] Among them, the input data of the first waveform generator 301 and the second waveform generator 302 are directly set by the computer control system 8. The two output ports of the first 90° electrical mixer 305 are respectively connected to the I and Q electrical input ports of the first IQ modulator 307; the two output ports of the second 90° electrical mixer 306 are respectively connected to the I and Q electrical input ports of the second IQ modulator 308, so that the data input by the computer control system 8 is finally modulated onto the optical signal after passing through the modulator driver and the electrical mixer.

[0081] In some embodiments, the second optical delay device 4 includes:

[0082] A second optical fiber amplifier 402, the input end of the second optical fiber amplifier 402 is connected to the output end of the first IQ modulator 307, and the output end of the second optical fiber amplifier 402 is connected to the input end of the optical mixer 5. Preferably, the second optical fiber amplifier 402 is a polarization-maintaining erbium-doped optical fiber amplifier.

[0083] A second tunable optical delay line 401, its input end is connected to the output end of the second IQ modulator 308. The second tunable optical delay line 401 is set by the computer control system 8 to ensure that the two signal optical paths are equal.

[0084] A third optical fiber amplifier 403, the input end of the third optical fiber amplifier 403 is connected to the output end of the second tunable optical delay line 401, and the output end of the third optical fiber amplifier 403 is connected to the input end of the optical mixer 5.

[0085] Specifically, the output port of the second optical fiber amplifier 402 is connected to the input port ① of the optical mixer 5; the output port of the third optical fiber amplifier 403 is connected to the input port ② of the optical mixer 5. The two input optical signals are mixed in the 180° optical mixer 5 to generate two optical outputs ③ and ④.

[0086] The output port ③ of the optical mixer 5 is connected to an input port of the photoelectric balanced detector 6; the output port ④ is connected to the other input port of the photoelectric balanced detector 6. The optical signal is converted into an electrical signal by the photoelectric balanced detector 6. The output port of the photoelectric balanced detector 6 is connected to the input port of the data acquisition device 7. The output port of the data acquisition device 7 is connected to the computer control system 8.

[0087] The following further illustrates with a specific example:

[0088] (1) The central wavelength of the laser source 1 is set to 1550.118 nm, and the power is set to 0 dBm. To ensure that the input power of the photoelectric balanced detector 6 is strong enough, the power gain of each optical amplifier is set to 23 dB.

[0089] (2) Use the computer control system 8 to set the first adjustable optical delay line 203 to ensure that the two beams of light are in phase.

[0090] (3) Use the computer control system 8 to input data into the first waveform generator 301 and the second waveform generator 302. The first waveform generator 301 and the second waveform generator 302 generate corresponding electrical signals according to the input data. The two electrical signals are amplified by the first modulator driver 303 and the second modulator driver 304 respectively, and are input into the first 90° electrical mixer 305 and the second 90° electrical mixer 306 respectively. The generated I and Q electrical signals are input into the electrical modulation input ports of the first IQ modulator 307 and the second IQ modulator 308 respectively, and finally the input data is loaded onto the optical carrier.

[0091] (4) Use the control system 8 to set the second adjustable optical delay line 401 to ensure that the two beams of light have equal optical path lengths. The two beams of light with equal optical path lengths obtained are input into the second optical fiber amplifier 402 and the third optical fiber amplifier 403 respectively, and the output gains of the second optical fiber amplifier 402 and the third optical fiber amplifier 403 are set to 23 dB through the control system 8.

[0092] (5) Then, input the two beams of light signals into the optical mixer 5 through the input ports ① and ②. After passing through the 180° optical mixer 5, the two beams of light signals generate optical mixing signals with phase differences of 0° and 180° at the output ports ③ and ④ respectively.

[0093] (6) Input the obtained optical mixing signals with phase differences of 0° and 180° into the optoelectronic balanced detector 6; the current generated by the optoelectronic balanced detector 6 is collected by the data acquisition device 7, and the convolution result is determined according to the collected electrical signal, and finally uploaded to the computer control system 8 for feedback to obtain the final calculation result.

[0094] It should be noted that the mathematical principle in the present invention is as follows:

[0095] Assume that the data of the tensor to be input is input at the first waveform generator 301 and is defined as where A i is the data of the input tensor, and the data of the weight tensor is input at the second waveform generator 302 and is defined as where, B k is the data of the weight tensor, and there is

[0096] Assume that the modulated single-wavelength narrow-linewidth laser source is C cos ω c t, where C is the amplitude of the light source. The output signals of the first IQ modulator 307 and the second IQ modulator 308 can be respectively expressed as

[0097]

[0098]

[0099] After the above two signals are received by input terminals ① and ② of the 180° optical mixer 5 respectively, optical mixing signals E1(t) and E2(t) with phase differences of 0° and 180° respectively are generated at output ports ③ and ④.

[0100] E1(t) = M s1 (t) + M s2 (t) (3)

[0101] E2(t) = M s1 (t) - M s2 (t) (4)

[0102] Subsequently, the optical mixing signals E1(t) and E2(t) generated at output ports ③ and ④ of the 180° optical bridge 5 are respectively received by the two input ports of the photoelectric balanced detector 6 for detection. The output electrical signals of the photoelectric balanced detector 6 are I1(t) and I2(t) respectively, where α is the photoelectric conversion efficiency, which is determined by the performance of the photodetector.

[0103]

[0104] Due to the limited working bandwidth of the detector, the actual current output is

[0105]

[0106] A band-pass filter can be used to filter out the signal with frequency and the amplitude can be read to obtain It should be noted that z ∈ Z, otherwise image frequency superposition will occur, that is, there will be superposition of the amplitudes of the waveforms with frequency, resulting in invalid results.

[0107] The electrical signal is collected by the data acquisition device 7 and output to the computer control system 8.

[0108] So far, the convolution result can be given by formula (6).

[0109] Among them, the length N of the input tensor and the length M of the weight tensor are both adjustable. Therefore, arbitrary convolution operations can be performed at the computer control system level (N and M are restricted by the orthogonal frequency interval of the electrical signal and the working bandwidths of optical and electrical devices), realizing adjustable configuration.

[0110] In summary, in the optoelectronic hybrid tensor convolution computing system of the present invention, the acquired original optical carrier is split and adjusted into two optical carriers with the same frequency and phase; and the input information is loaded onto the two optical carriers and adjusted into two optical signals with equal optical paths. A complete convolution result can be obtained through one operation by the balanced optoelectronic detector. Moreover, the parameters of each adjustable device can be modified through the computer control system, which is convenient to set and simple and convenient. And each optical device adopts a fiber device and can be integrated on-chip subsequently. Therefore, it has the advantages of high signal-to-noise ratio, large throughput, and easy subsequent on-chip integration, and can well meet the requirements of the optical computing field.

[0111] The above is only the specific implementation manner of the embodiments of the present invention, but the protection scope of the embodiments of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the embodiments of the present invention can easily think of various equivalent modifications or substitutions, and these modifications or substitutions should all be covered within the protection scope of the embodiments of the present invention. Therefore, the protection scope of the embodiments of the present invention should be subject to the protection scope of the claims.

Claims

1. An optoelectronic hybrid tensor convolution calculation system, characterized in that Comprising: A laser source for generating an original optical carrier; A first optical delay and beam splitter device for splitting and adjusting the original optical carrier into two optical carriers with the same frequency and phase, and the first optical delay and beam splitter device includes a beam splitter and a first tunable optical delay line; An electro-optic modulation device for loading input information onto the two optical carriers; A second optical delay device for adjusting the two modulated optical carriers into two optical signals with equal optical paths; An optical mixer for mixing the two optical signals to generate two optical mixing signals with phase differences of 0° and 180° respectively; An optoelectronic balanced detector for performing optoelectronic conversion on the two optical mixing signals to output an electrical signal; A data acquisition device for acquiring the output electrical signal and determining the convolution result based on the amplitude of the acquired electrical signal after being filtered by a filter with a corresponding frequency; A computer control system for controlling the adjustment parameters of the electro-optic modulation device and the second tunable optical delay line; The electro-optic modulation device includes: A first modulation component including a first waveform generator, a first modulator driver, a first 90° electrical mixer, and a first IQ modulator connected in sequence, and the input end of the first IQ modulator is directly connected to the output end of the beam splitter; A second modulation component including a second waveform generator, a second modulator driver, a second 90° electrical mixer, and a second IQ modulator connected in sequence, and the input end of the second IQ modulator is connected to the output end of the first tunable optical delay line; The electro-optic modulation device loading the input information onto the two optical carriers includes: The electro-optic modulation device generates two corresponding analog electrical signals according to the input data, generates two electrical drive signals with a phase difference of 90° based on the two analog electrical signals, and loads the input data onto the two optical carriers according to the electrical drive signals and the corresponding optical carriers; The second optical delay device adjusting the two modulated optical carriers into two optical signals with equal optical paths includes: The second optical delay device adjusts the optical paths of the modulated optical carriers to generate two optical signals with equal optical paths.

2. The computing system for optoelectronic hybrid tensor convolution according to claim 1, characterized in that, The first optical delay and beam splitter device includes: A beam splitter for splitting the original optical carrier into two optical carriers; A first tunable optical delay line with its input end connected to the beam splitter for adjusting the optical path of the optical carrier.

3. The computing system for optoelectronic hybrid tensor convolution according to claim 2, wherein The first optical delay and beam splitter device further includes: A first fiber amplifier with its input end connected to the output end of the laser source and its output end connected to the input end of the beam splitter.

4. The computing system for optoelectronic hybrid tensor convolution according to claim 1, wherein The second optical delay device includes: A second fiber amplifier with its input end connected to the output end of the first IQ modulator and its output end connected to the input end of the optical mixer; A second tunable optical delay line with its input end connected to the output end of the second IQ modulator; A third fiber amplifier with its input end connected to the output end of the second tunable optical delay line and its output end connected to the input end of the optical mixer.

5. The computing system for optoelectronic hybrid tensor convolution according to claim 4, characterized in that: The computer control system is also used to control the adjustment parameters of the first tunable optical delay line.

6. The computing system for optoelectronic hybrid tensor convolution according to claim 5, characterized in that: The data acquisition device is used to feedback the convolution result to the computer control system.

Images