Microwave photon neuron convolution acceleration device for electromagnetic signal processing

Abstract

The invention discloses an electromagnetic signal processing-oriented microwave photon neuron convolution acceleration device, which relates to the technical field of signal processing, and comprises a microwave photon neuron convolution accelerator, an electromagnetic signal processing main control unit, a signal processing main control unit and a signal processing main control unit, the matrix multiplication and addition linear operation unit is used for completing matrix multiplication and addition linear operation of a received electromagnetic signal or processing data and a preset weight vector on an optical simulation domain, and outputting a matrix multiplication and addition linear operation result to the electromagnetic signal processing main control unit; the electromagnetic signal processing main control unit is used for inputting a pre-obtained electromagnetic signal in a vector format to the microwave photon neuron convolution accelerator, processing the received matrix multiplication and addition linear operation result to obtain processed data in the vector format, and outputting the processed data in the vector format to the microwave photon neuron convolution accelerator; and outputting the processed data to the microwave photon neuron convolution accelerator when the preset condition is not met. According to the invention, rapid processing and identification of the broadband electromagnetic signal can be realized, the calculation energy consumption demand is reduced, and the calculation power demand and the energy efficiency ratio demand of an electronic device are reduced.

Description

Technical Field

[0001] The present invention relates to the field of signal processing technology, and in particular to a microwave photon neuron convolution acceleration device for electromagnetic signal processing. Background Art

[0002] With the development and large-scale application of emerging technologies such as the Internet of Things, video communications, satellite communications, and radar, the use of electromagnetic spectrum resources has become very important in various fields. Although the International Telecommunication Union (ITU) has divided the use of frequency bands, electromagnetic interference or illegal use between different services may still cause damage to important electromagnetic systems. In addition, with the development of the above-mentioned emerging technologies, the requirements for signal bandwidth and real-time processing of large-scale signals are becoming higher and higher. Therefore, high-performance electromagnetic signal receiving and processing equipment is of great significance for the perception and monitoring of electromagnetic spectrum resources. Electromagnetic signal receiving and processing equipment of traditional microwave / digital signal systems is usually limited by the analog-to-digital conversion (ADC) and electronic component performance of the digital signal processing part. It is difficult to achieve high-speed processing and real-time perception of large-bandwidth signals under the condition of limited cost and power consumption resources.

[0003] The high parallelism and high processing speed of optical computing technology can quickly process large amounts of data to meet the signal processing requirements of modern communications, radar, remote sensing, electronic reconnaissance and other systems; secondly, optical computing technology can realize nonlinear signal processing, and accelerate the calculation of neural network models by constructing single or multiple neuron models and nonlinear activation functions, which is particularly suitable for the intelligent and rapid identification of electromagnetic targets; in addition, combined with the large bandwidth characteristics of microwave photons, it can cover multiple frequency bands, which is conducive to the classification and identification of electromagnetic signals under multiple scene changes; and optical computing technology has good anti-electromagnetic interference capabilities, which is also very beneficial for the normal operation and stable operation of devices in the field of electromagnetic signal processing. For this reason, microwave photon-optical computing technology provides a new idea for electromagnetic signal perception and processing.

[0004] At present, optical neural networks and photonic convolution calculations based on microwave photonic optoelectronic hybrid processing architectures have been widely studied and applied. For example, Xingyuan Xu, Mengxi Tan, et al. published a paper titled "11 TOPS photonic convolutional accelerator for optical neural networks" in the journal Nature in 2021, proposing an optical vector convolution accelerator that can run at a speed of more than 10TOPS (trillion operations per second) and can be used for large-scale image data processing and rapid image recognition. Wolfram Pernice et al. published a paper titled "Integrated PhotonicTensor Core Based on Phase-Change Memory Arrays" in the journal Nature Communications in 2021, proposing an integrated photonic tensor core based on a phase-change material memory array that can run at a speed of about 0.5TOPS (trillion operations per second), realizes parallel matrix-vector multiplication operations, and tests image recognition effects on a handwritten digit database.

[0005] At present, the main research is on the application of microwave photon-optical computing technology in the field of image processing and recognition tasks. The acceleration devices used for image processing and recognition tasks cannot be applied to the rapid processing and recognition of electromagnetic signals. Moreover, the performance of the current acceleration devices based on microwave photon-optical computing technology is highly dependent on the performance of optical materials, the device structure is relatively complex, the cost of setting up the device is high, and there are great limitations. Summary of the invention

[0006] In order to solve some or all of the technical problems existing in the above-mentioned prior art, the present invention provides a microwave photon neuron convolution acceleration device for electromagnetic signal processing.

[0007] The technical solution of the present invention is as follows: A microwave photon neuron convolution acceleration device for electromagnetic signal processing is provided, comprising: A microwave photon neuron convolution accelerator, connected to the electromagnetic signal processing main control unit, is used to complete the matrix multiplication and addition linear operation of the received electromagnetic signal or processing data and the preset weight vector in the optical simulation domain, and output the matrix multiplication and addition linear operation result to the electromagnetic signal processing main control unit; The electromagnetic signal processing main control unit is used to input a pre-obtained electromagnetic signal in a vector format into the microwave photonic neuron convolution accelerator, and to process the received matrix multiplication and addition linear operation results to obtain processed data in a vector format, and output the processed data to the microwave photonic neuron convolution accelerator when the preset conditions are not met.

[0008] In some optional embodiments, the microwave photon neuron convolution accelerator comprises: A multi-wavelength optical excitation unit is connected to the weight control unit, and is used to generate multiple optical carriers with different wavelengths and equal intervals between adjacent wavelengths, and synthesize the multiple optical carriers into one optical carrier and output it to the weight control unit; The weight control unit is connected to the optoelectronic modulation unit. The weight control unit is preset with a weight vector, which is used to load weights on the received optical carrier according to the weight vector, and divide the optical carrier into multiple channels according to different wavelengths, and output the multiple optical carriers in parallel to the optoelectronic modulation unit, wherein the optical carrier of each wavelength is loaded with the same weight; An input vector processing unit, connected to the photoelectric modulation unit, and used to output the electromagnetic signal or processed data sent by the electromagnetic signal processing main control unit to the photoelectric modulation unit; The optoelectronic modulation unit is connected to the wavelength delay unit, and is used to modulate the received electromagnetic signal or processed data onto the multi-channel optical carriers input in parallel by the weight control unit to obtain multi-channel optical signals, and output the multi-channel optical signals in parallel to the wavelength delay unit, wherein each wavelength optical carrier is loaded with the same modulation signal, and the modulation signal is the electromagnetic signal or processed data received by the optoelectronic modulation unit; The wavelength delay unit is connected to the feature extraction unit, and is used to apply time delay to optical signals of different wavelengths, so that multiple optical signals are delayed in sequence by a set time length according to the size of the wavelength; The feature extraction unit is connected to the photoelectric detection unit, and is used to extract the light signal fragment within the set time window from the delayed light signal and output it to the photoelectric detection unit; The photoelectric detection unit is used to linearly superimpose the received multi-path optical signals.

[0009] In some optional implementations, the set duration is expressed as: ; in, Indicates the set duration. It indicates the modulation period required for the optoelectronic modulation unit to modulate the modulation signal to an optical carrier. Indicates the number of wavelength types of the optical signal.

[0010] In some optional embodiments, the multi-wavelength optical excitation unit includes an optical frequency comb.

[0011] In some optional embodiments, the weight control unit includes a waveform shaper and an optical beam splitter connected in sequence.

[0012] In some optional embodiments, the electro-optical modulation unit includes a broadband electro-optical modulator.

[0013] In some optional embodiments, the wavelength delay unit includes a dispersive optical fiber.

[0014] In some optional embodiments, the photodetection unit comprises a broadband flat intensity detector.

[0015] In some optional embodiments, the microwave photon neuron convolution accelerator includes two input ports and one output port, and the electromagnetic signal processing main control unit includes one input port and two output ports; The two input ports of the microwave photon neuron convolution accelerator are respectively connected to the two output ports of the electromagnetic signal processing main control unit, and the output port of the microwave photon neuron convolution accelerator is connected to the input port of the electromagnetic signal processing main control unit. The electromagnetic signal processing main control unit inputs electromagnetic signals and processing data to the microwave photon neuron convolution accelerator through the two output ports respectively, and the microwave photon neuron convolution accelerator inputs matrix multiplication and addition linear operation results to the electromagnetic signal processing main control unit through the output port.

[0016] In some optional implementations, it also includes a master control host computer; The main control host computer is connected to the electromagnetic signal processing main control unit. The electromagnetic signal processing main control unit outputs processed data to the main control host computer when a preset condition is met. The main control host computer is used to display the received processed data.

[0017] The main advantages of the technical solution of the present invention are as follows: The microwave photon neuron convolution acceleration device of the present invention utilizes a microwave photon neuron convolution accelerator to migrate large-scale matrix multiplication and addition linear operations to the optical simulation domain, which can significantly reduce the demand for computing energy consumption of electronic devices, and reduce the demand for computing power and energy efficiency ratio of electronic devices; by utilizing the advantages of the microwave photon neuron convolution accelerator's high parallelism, high speed, and large bandwidth, it can achieve rapid processing and recognition of broadband electromagnetic signals. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The drawings described herein are used to provide a further understanding of the embodiments of the present invention and constitute a part of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings: Figure 1 A schematic diagram of the structure of a microwave photon neuron convolution acceleration device for electromagnetic signal processing provided by an embodiment of the present invention; Figure 2 A schematic diagram of the structure of a microwave photon neuron convolution accelerator provided by an embodiment of the present invention; Figure 3 A schematic diagram of the principle of implementing matrix multiplication and addition linear operations by a microwave photon neuron convolution accelerator provided by an embodiment of the present invention; Figure 4 A schematic diagram of the structure of another microwave photon neuron convolution accelerator provided by an embodiment of the present invention, wherein the feature extraction unit is not shown; Figure 5 A schematic diagram of the operation process of implementing matrix multiplication and addition linear operations on the time-wavelength plane for 32 parallel optical signals provided in an embodiment of the present invention; Figure 6 A schematic diagram of the structure of another microwave photon neuron convolution acceleration device for electromagnetic signal processing provided by an embodiment of the present invention; Figure 7 A schematic diagram of the recognition accuracy of 8 electromagnetic signals provided in Example 1 of the present invention. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the specific embodiments of the present invention and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0020] The technical solution provided by the embodiments of the present invention is described in detail below with reference to the accompanying drawings.

[0021] refer to Figure 1 The embodiment of the present invention provides a microwave photon neuron convolution acceleration device for electromagnetic signal processing, the device comprising: The microwave photon neuron convolution accelerator 1 is connected to the electromagnetic signal processing main control unit 2, and is used to complete the matrix multiplication and addition linear operation (MM-MAC) of the received electromagnetic signal or processing data and the preset weight vector in the optical simulation domain, and output the matrix multiplication and addition linear operation result to the electromagnetic signal processing main control unit 2; The electromagnetic signal processing main control unit 2 is used to input a pre-obtained electromagnetic signal in a vector format to the microwave photon neuron convolution accelerator 1, and to process the received matrix multiplication and addition linear operation results to obtain processed data in a vector format, and output the processed data to the microwave photon neuron convolution accelerator 1 when the preset conditions are not met.

[0022] Specifically, when the microwave photon neuron convolution acceleration device provided by the embodiment of the present invention is used, the electromagnetic signal to be identified and processed is first sent to the electromagnetic signal processing main control unit 2, and the electromagnetic signal processing main control unit 2 converts the received electromagnetic signal into a vector format and sends it to the microwave photon neuron convolution accelerator 1. The electromagnetic signal in the vector format can be regarded as the input of the first layer of neurons in the neural network; the microwave photon neuron convolution accelerator 1 completes the matrix multiplication and linear operation of the received electromagnetic signal and the preset weight vector in the optical simulation domain, and then outputs the obtained matrix multiplication and linear operation result to the electromagnetic signal processing main control unit 2; the electromagnetic signal processing main control unit 2 performs preset data post-processing on the received matrix multiplication and linear operation to obtain processed data in a vector format, and the processed data can be regarded as the output of the first layer of neurons in the neural network; after obtaining the processed data, the electromagnetic signal processing main control unit Unit 2 determines whether the preset conditions are met. If not, the processed data is output to the microwave photon neuron convolution accelerator 1. The processed data in vector format can be regarded as the input of the second layer of neurons in the neural network. The microwave photon neuron convolution accelerator 1 completes the matrix multiplication and linear operation of the received processed data and the preset weight vector in the optical simulation domain, and then outputs the obtained matrix multiplication and linear operation result to the electromagnetic signal processing main control unit 2. The electromagnetic signal processing main control unit 2 performs preset data post-processing on the received matrix multiplication and linear operation to obtain processed data in vector format. The processed data can be regarded as the output of the second layer of neurons in the neural network. Then, it continues to determine whether the preset conditions are met. If not, the processing continues according to the above-mentioned processing process of the processed data until the preset conditions are met, and the finally obtained processed data is used as the electromagnetic signal processing result.

[0023] In an embodiment of the present invention, by taking the electromagnetic signal to be identified and processed as the initial input, the matrix multiplication and addition linear operation processing of the microwave photon neuron convolution accelerator 1 and the data post-processing of the electromagnetic signal processing main control unit 2 are repeated multiple times, so that the multi-layer neural network can realize the processing and identification process of the electromagnetic signal, thereby completing the rapid processing and identification of the electromagnetic signal.

[0024] In an embodiment of the present invention, the preset conditions in the electromagnetic signal processing main control unit 2 are specifically set according to actual conditions. For example, the preset condition is that the number of times the electromagnetic signal processing main control unit 2 performs data post-processing reaches a set threshold value M. At this time, the matrix multiplication and addition linear operation processing of the microwave photon neuron convolution accelerator 1 and the data post-processing of the electromagnetic signal processing main control unit 2 can be cyclically performed M times, thereby forming a neural network with M layers of neurons to realize the processing and identification of electromagnetic signals.

[0025] In the embodiment of the present invention, the electromagnetic signal processing main control unit 2 performs data post-processing on the matrix multiplication and addition linear operation according to the actual situation. For example, the data post-processing includes: sampling, quantization, biasing and nonlinear activation processing performed in sequence.

[0026] The microwave photon neuron convolution acceleration device provided in the embodiment of the present invention utilizes the microwave photon neuron convolution accelerator 1 to migrate large-scale matrix multiplication and addition linear operations to the optical simulation domain, which can greatly reduce the demand for computing energy consumption of electronic devices, and reduce the demand for computing power and energy efficiency ratio of electronic devices; by utilizing the advantages of the microwave photon neuron convolution accelerator 1 of high parallelism, high speed and large bandwidth, it can realize rapid processing and recognition of broadband electromagnetic signals.

[0027] refer to Figure 2 In an embodiment of the present invention, in order to realize the functions of the microwave photon neuron convolution accelerator 1 defined above, reduce the structural complexity of the microwave photon neuron convolution accelerator 1, and facilitate the implementation of the microwave photon neuron convolution accelerator 1, the microwave photon neuron convolution accelerator 1 includes: The multi-wavelength optical excitation unit 11 is connected to the weight control unit 12. The multi-wavelength optical excitation unit 11 is used to generate multiple optical carriers with different wavelengths and equal intervals between adjacent wavelengths, and synthesize the multiple optical carriers into one and output them to the weight control unit 12; The weight control unit 12 is connected to the optoelectronic modulation unit 14. The weight control unit 12 is preset with a weight vector, which is used to load weights on the received optical carrier according to the weight vector, and divide the optical carrier into multiple channels according to different wavelengths, and output the multiple optical carriers in parallel to the optoelectronic modulation unit 14, wherein the optical carrier of each wavelength is loaded with the same weight; The input vector processing unit 13 is connected to the photoelectric modulation unit 14 and is used to output the electromagnetic signal or the processed data sent by the electromagnetic signal processing main control unit 2 to the photoelectric modulation unit 14; The photoelectric modulation unit 14 is connected to the wavelength delay unit 15. The photoelectric modulation unit 14 is used to modulate the received electromagnetic signal or processed data onto the multi-channel optical carriers input in parallel by the weight control unit 12 to obtain the multi-channel optical signals, and output the multi-channel optical signals in parallel to the wavelength delay unit 15, wherein each wavelength optical carrier is loaded with the same modulation signal, and the modulation signal is the electromagnetic signal or processed data received by the photoelectric modulation unit 14; The wavelength delay unit 15 is connected to the feature extraction unit 16, and is used to apply time delay to optical signals of different wavelengths, so that multiple optical signals are delayed in sequence according to the wavelength size for a set time length; The feature extraction unit 16 is connected to the photoelectric detection unit 17, and the feature extraction unit 16 is used to extract the light signal fragment within the set time window from the delayed light signal and output it to the photoelectric detection unit 17; The photoelectric detection unit 17 is used to linearly superimpose the received multi-path optical signals.

[0028] In an embodiment of the present invention, the multi-wavelength optical excitation unit 11 generates a plurality of parallel optical carriers with equal wavelength intervals to construct a time-wavelength plane. The multi-wavelength optical excitation unit 11 provides multiple parallel optical carriers as a calculation carrier, each wavelength corresponds to an independent signal channel, and outputs all wavelengths at the initial moment to form a signal plane interlaced with the time dimension and the wavelength dimension, providing a physical basis for the subsequent loading of weights and modulation signals. The weight control unit 12 loads the same weight vector for the optical carrier of each wavelength. The weight control unit 12 maps the weight information to the optical carrier by means such as optical power adjustment or wavelength shift, so as to store the weight information on the time-wavelength plane and realize the pre-association of the weight and the modulation signal. The input vector processing unit 13 is used as a modulation signal source to provide a modulation signal. The input vector processing unit 13 cooperates with the optoelectronic modulation unit 14 to ensure the accurate mapping of the modulation signal on the time-wavelength plane. The optoelectronic modulation unit 14 is used to uniformly load the modulation signal provided by the input vector processing unit 13 to all wavelengths, so that the optical carrier of each wavelength carries the same modulation signal, and ensures the segmented alignment of the signal on the time-wavelength plane. The wavelength delay unit 15 introduces a specific delay to the optical signals of different wavelengths to adjust the timing of the optical signals, specifically stretching the diagonal elements on the time-wavelength plane to the same time window to provide timing alignment for subsequent feature extraction and accumulation. The feature extraction unit 16 extracts the smallest feature unit from the delayed optical signal, such as the optical signal fragment within a specific time window, completes the segmented cutting of the signal, and ensures the independence of subsequent accumulation operations. The photoelectric detection unit 17 converts the optical signal extracted by the feature extraction unit 16 into an electrical signal, and through the integration effect, completes the linear superposition of all signals within a specific time window to obtain the matrix multiplication and linear operation result of the modulated signal and the weight vector.

[0029] In an embodiment of the present invention, the photoelectric modulation unit 14, the wavelength delay unit 15, the feature extraction unit 16 and the photoelectric detection unit 17 each include one or more, and the number of the photoelectric modulation unit 14, the wavelength delay unit 15, the feature extraction unit 16 and the photoelectric detection unit 17 is the same. Wherein, when the photoelectric modulation unit 14, the wavelength delay unit 15, the feature extraction unit 16 and the photoelectric detection unit 17 are all one, the microwave photon neuron convolution accelerator 1 can constitute a one-dimensional microwave photon neuron; when the photoelectric modulation unit 14, the wavelength delay unit 15, the feature extraction unit 16 and the photoelectric detection unit 17 are all multiple, the microwave photon neuron convolution accelerator 1 can constitute a multi-dimensional microwave photon neuron. For example, when the photoelectric modulation unit 14, the wavelength delay unit 15, the feature extraction unit 16 and the photoelectric detection unit 17 are all N, the microwave photon neuron convolution accelerator 1 constitutes an N-dimensional microwave photon neuron, and N is a preset positive integer.

[0030] It is assumed that the multi-wavelength optical excitation unit 11 generates n wavelengths respectively: The weight vector preset in the weight control unit 12 is , the signal vector output by the input vector processing unit 13 is , the principle of matrix multiplication and addition linear operation of a single neuron in the microwave photon neuron convolution accelerator 1, which is composed of a multi-wavelength optical excitation unit 11, a weight control unit 12, an input vector processing unit 13, an optoelectronic modulation unit 14, a wavelength delay unit 15, a feature extraction unit 16 and a photoelectric detection unit 17, is as follows: Figure 3 Specifically, the multi-wavelength optical excitation unit 11 generates n wavelengths, which are The n optical carriers are combined into one output to construct an initial time-wavelength plane; after the n optical carriers pass through the weight control unit 12, the optical carriers of each wavelength are loaded with the same weight vector After the optical carrier loaded with the weight vector is output to the optoelectronic modulation unit 14, the optoelectronic modulation unit 14 inputs the signal vector output by the vector processing unit 13 As the modulation signal, it is uniformly modulated onto n optical carriers, so that each wavelength of the optical carrier is loaded with the same modulation signal. ; When the modulation period is When the signal vector The duration of each element in is ; Through the processing of the photoelectric modulation unit 14, the weight vector With the signal vector Complete linear weighting, weight information and signal information are stored in the time-wavelength plane; then use the wavelength delay unit 15 to delay the time-wavelength plane containing the weight information, so that the same amount of delay is generated between each wavelength on the time-wavelength plane. The specific delay duration is The time-wavelength plane containing weight information is stretched in the time dimension of the plane after the delay of the wavelength delay unit 15, by making the delay between different wavelengths , which can make the elements on the diagonal of the time-wavelength plane containing weight information Stretched to the same time window The feature extraction unit 16 extracts and divides the elements of the same time window into the smallest feature unit, and outputs it to the photoelectric detection unit 17; the photoelectric detection unit 17 has no wavelength selectivity and has an optical signal amplitude response characteristic, which can correspond to the elements in the time window. Extract all of them, complete the accumulation operation of each weighted component, and get the final operation result , .

[0031] refer to Figure 4 In order to facilitate the realization of the microwave photonic neuron convolution accelerator 1, in an embodiment of the present invention, the multi-wavelength optical excitation unit 11 includes an optical frequency comb, the weight control unit 12 includes a waveform shaper and an optical beam splitter, the photoelectric modulation unit 14 includes a broadband photoelectric modulator, the wavelength delay unit 15 includes a dispersion optical fiber, and the photoelectric detection unit 17 includes a broadband flat intensity detector.

[0032] The waveform shaper is connected to the optical beam splitter, the waveform shaper is used for weight loading, and the optical beam splitter is used for dividing the optical signal after the weight loading into multiple outputs.

[0033] In the embodiment of the present invention, by adopting the above-mentioned optical devices and electrical devices to construct a microwave photon neuron convolution accelerator 1, the practicability of the microwave photon neuron convolution accelerator can be improved and the cost of the device can be reduced.

[0034] refer to Figure 4 Furthermore, considering that the output optical power of the optical frequency comb is low, the multi-wavelength optical excitation unit 11 also includes an optical amplifier, which is connected to the optical frequency comb and is used to amplify the power of the optical signal output by the optical frequency comb.

[0035] In the embodiment of the present invention, when the wavelength delay unit 15 includes a dispersive optical fiber, the delay control of the optical signal is achieved by utilizing the dispersion control of the dispersive optical fiber, and the corresponding delay duration is expressed as: ; in, Indicates the set duration. Indicates the dispersion coefficient of the dispersive fiber, in ps / km·nm. Represents the wavelength interval of the optical signal, Indicates the fiber length of dispersive fiber.

[0036] According to the time length expression corresponding to the above dispersive optical fiber, when the wavelength remains unchanged, the delay time length can be adjusted by adjusting the dispersion coefficient and the optical fiber length of the dispersive optical fiber.

[0037] Furthermore, in the embodiment of the present invention, the broadband electro-optical modulator may be an electro-optical modulator with a 3dB bandwidth > 20 GHz, and the broadband flat intensity detector may be a flat intensity detector with a bandwidth > 10 GHz.

[0038] refer to Figure 5 , Figure 5 The 32-channel parallel optical signal is shown based on the microwave photon neuron convolution accelerator 1 provided in the embodiment of the present invention. The operation process of matrix multiplication and addition linear operation is implemented on the time-wavelength plane. Assume that the input vector is [0,1,0,0], the clock period of the input vector is 32 times the delay of the adjacent optical signal, and the weight vector is 32 bits [1,1,…,1]. Figure 4 The dual-path parallel microwave photonic neuron convolution accelerator shown can realize matrix multiplication and addition linear operations.

[0039] refer to Figure 1 ,Further, in the embodiment of the present invention, the microwave photon neuron convolution accelerator 1 includes two input ports and one output port, and the electromagnetic signal processing main control unit 2 includes one input port and two output ports; The two input ports of the microwave photon neuron convolution accelerator 1 are respectively connected to the two output ports of the electromagnetic signal processing main control unit 2, and the output port of the microwave photon neuron convolution accelerator 1 is connected to the input port of the electromagnetic signal processing main control unit 2. The electromagnetic signal processing main control unit 2 inputs electromagnetic signals and processing data to the microwave photon neuron convolution accelerator 1 through the two output ports, and the microwave photon neuron convolution accelerator 1 inputs matrix multiplication and addition linear operation results to the electromagnetic signal processing main control unit 2 through the output port.

[0040] In the embodiment of the present invention, by setting two input ports on the microwave photonic neuron convolution accelerator 1 and setting two output ports on the electromagnetic signal processing main control unit 2, the two output ports are used to input electromagnetic signals and process data into the microwave photonic neuron convolution accelerator 1 respectively, so as to avoid interference when transmitting electromagnetic signals and processing data, and ensure the accuracy of data processing results.

[0041] refer to Figure 6 , Further, in the embodiment of the present invention, the microwave photon neuron convolution acceleration device also includes a main control host computer 3; The main control host computer 3 is connected to the electromagnetic signal processing main control unit 2. The electromagnetic signal processing main control unit 2 outputs the processed data to the main control host computer 3 when the preset conditions are met. The main control host computer 3 is used to display the received processed data.

[0042] The main control host computer 3 can also regulate and control the electromagnetic signal processing main control unit 2 , specifically, it can input the electromagnetic signal to be identified and processed into the electromagnetic signal processing main control unit 2 , and set preset conditions in the electromagnetic signal processing main control unit 2 .

[0043] By setting up the main control host computer 3, it is possible to facilitate the adjustment and control of the electromagnetic signal processing main control unit 2, and to facilitate the output and display of the final processed data, so as to facilitate subsequent calls.

[0044] Furthermore, in the embodiment of the present invention, the electromagnetic signal processing main control unit 2 is constructed using a digital-to-analog converter (DAC), an analog-to-digital converter (ADC), a field programmable gate array (FPGA) and a graphics processing unit (GPU). The main control host computer 3 selects an X86 system or an Ubuntu system and supports 32-bit or 64-bit floating point calculations.

[0045] The beneficial effects of the microwave photon neuron convolution acceleration device for electromagnetic signal processing provided by the embodiment of the present invention are described below with reference to specific examples: Example 1 In this example 1, build Figure 4 The dual-path parallel microwave photon neuron convolution accelerator shown in the figure performs matrix multiplication and addition linear operations on electromagnetic signals, and builds an electromagnetic signal processing main control unit 2 to collect, quantize, bias and perform nonlinear activation processing on the results of matrix multiplication and addition linear operations. Eight types of electromagnetic signals including AM-DSB signal, AM-SSB signal, BPSK signal, CPFSK signal, GFSK signal, 4-PAM signal, 16-QAM signal and QPSK signal are selected as electromagnetic signals to be identified and processed. The electromagnetic signals to be identified and processed are identified and processed using the built microwave photon neuron convolution accelerator 1 and the electromagnetic signal processing main control unit 2, and the following is obtained: Figure 7 It can be seen that the microwave photon neuron convolution acceleration device for electromagnetic signal processing provided by the embodiment of the present invention can complete the rapid processing and recognition of electromagnetic signals.

[0046] It should be noted that, in this article, relational terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In addition, "front", "back", "left", "right", "upper" and "lower" in this article are all referenced to the placement state shown in the accompanying drawings.

[0047] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A microwave photon neuron convolution accelerator for electromagnetic signal processing, characterized in that: include: A microwave photon neuron convolution accelerator, connected to the electromagnetic signal processing main control unit, is used to complete the matrix multiplication and addition linear operation of the received electromagnetic signal or processing data and the preset weight vector in the optical simulation domain, and output the matrix multiplication and addition linear operation result to the electromagnetic signal processing main control unit; The electromagnetic signal processing main control unit is used to input a pre-obtained electromagnetic signal in a vector format into the microwave photonic neuron convolution accelerator, and to process the received matrix multiplication and addition linear operation results to obtain processed data in a vector format, and output the processed data to the microwave photonic neuron convolution accelerator when the preset conditions are not met.

2. The microwave photon neuron convolution accelerator for electromagnetic signal processing according to claim 1, characterized in that: The microwave photon neuron convolution accelerator comprises: A multi-wavelength optical excitation unit is connected to the weight control unit, and is used to generate multiple optical carriers with different wavelengths and equal intervals between adjacent wavelengths, and synthesize the multiple optical carriers into one optical carrier and output it to the weight control unit; The weight control unit is connected to the optoelectronic modulation unit. The weight control unit is preset with a weight vector, which is used to load weights on the received optical carrier according to the weight vector, and divide the optical carrier into multiple channels according to different wavelengths, and output the multiple optical carriers in parallel to the optoelectronic modulation unit, wherein the optical carrier of each wavelength is loaded with the same weight; An input vector processing unit, connected to the photoelectric modulation unit, and used to output the electromagnetic signal or processed data sent by the electromagnetic signal processing main control unit to the photoelectric modulation unit; The optoelectronic modulation unit is connected to the wavelength delay unit, and is used to modulate the received electromagnetic signal or processed data onto the multi-channel optical carriers input in parallel by the weight control unit to obtain multi-channel optical signals, and output the multi-channel optical signals in parallel to the wavelength delay unit, wherein each wavelength optical carrier is loaded with the same modulation signal, and the modulation signal is the electromagnetic signal or processed data received by the optoelectronic modulation unit; The wavelength delay unit is connected to the feature extraction unit, and is used to apply time delay to optical signals of different wavelengths, so that multiple optical signals are delayed in sequence by a set time length according to the size of the wavelength; The feature extraction unit is connected to the photoelectric detection unit, and is used to extract the light signal fragment within the set time window from the delayed light signal and output it to the photoelectric detection unit; The photoelectric detection unit is used to linearly superimpose the received multi-path optical signals.

3. The microwave photon neuron convolution acceleration device for electromagnetic signal processing according to claim 2 is characterized in that: The set duration is expressed as: ； in, Indicates the set duration. It indicates the modulation period required for the optoelectronic modulation unit to modulate the modulation signal to an optical carrier. Indicates the number of wavelength types of the optical signal.

4. The microwave photon neuron convolution acceleration device for electromagnetic signal processing according to claim 2 is characterized in that: The multi-wavelength optical excitation unit includes an optical frequency comb.

5. The microwave photon neuron convolution acceleration device for electromagnetic signal processing according to claim 2, characterized in that: The weight control unit includes a waveform shaper and an optical beam splitter connected in sequence.

6. The microwave photon neuron convolution acceleration device for electromagnetic signal processing according to claim 2, characterized in that: The optoelectronic modulation unit includes a broadband optoelectronic modulator.

7. The microwave photon neuron convolution acceleration device for electromagnetic signal processing according to claim 2, characterized in that: The wavelength delay unit includes a dispersive optical fiber.

8. The microwave photon neuron convolution acceleration device for electromagnetic signal processing according to claim 2, characterized in that: The photodetection unit includes a broadband flat intensity detector.

9. The microwave photon neuron convolution acceleration device for electromagnetic signal processing according to claim 1, characterized in that: The microwave photon neuron convolution accelerator includes two input ports and one output port, and the electromagnetic signal processing main control unit includes one input port and two output ports; The two input ports of the microwave photon neuron convolution accelerator are respectively connected to the two output ports of the electromagnetic signal processing main control unit, and the output port of the microwave photon neuron convolution accelerator is connected to the input port of the electromagnetic signal processing main control unit. The electromagnetic signal processing main control unit inputs electromagnetic signals and processing data to the microwave photon neuron convolution accelerator through the two output ports respectively, and the microwave photon neuron convolution accelerator inputs matrix multiplication and addition linear operation results to the electromagnetic signal processing main control unit through the output port.

10. The microwave photon neuron convolution accelerator for electromagnetic signal processing according to any one of claims 1 to 9, characterized in that: It also includes a master control host computer; The main control host computer is connected to the electromagnetic signal processing main control unit. The electromagnetic signal processing main control unit outputs processed data to the main control host computer when a preset condition is met. The main control host computer is used to display the received processed data.

Images