A memristor-based perceptual neuron circuit and application

Abstract

The application discloses a kind of perception neuron circuits based on memristor and application, belong to intelligent perception technical field;Integrated perception neuron array is constructed by word line and bit line, and each perception neuron in array is converted into pulse signal and is output in parallel with its perceived external environment information;Each perception neuron in array includes series resistance sensor and threshold transition type memristor;Sensor acts as adjustable resistance, so that memristor works in local active area and normally pulse firing, play the role of sensing environmental information and voltage division, also reduce the difference between devices and devices of multiple perception circuit integration memristor caused by encoding error of perception encoding circuit.In addition, the parasitic capacitance of memristor itself acts as necessary capacitor in memristor neuron circuit;Through the above design, the perception neuron circuit is more compact, greatly improves the integration density, and the reduction of area and hardware saving also reduces energy consumption, improves energy efficiency.

Description

Technical Field

[0001] This invention belongs to the field of intelligent sensing technology, and more specifically, relates to a sensing neuron circuit based on memristors and its application. Background Technology

[0002] With the continuous development of intelligent sensing computing, the construction of artificial intelligence robots places higher demands on hardware systems. For example, when robots work in complex environments (autonomous driving, intelligent fire fighting, deep-sea exploration), the ability to perceive and process information in real time is essential.

[0003] However, in traditional architectures, analog data collected by sensors is first converted into digital signals by an analog-to-digital converter, then stored in memory, and sent to the computing unit, resulting in high energy consumption and low efficiency. In contrast, neuromorphic perception computing systems inspired by biological senses offer advantages in processing perceptual information, including high energy efficiency, strong robustness, high flexibility, and low fault tolerance. However, current neuromorphic perception processing hardware technology is mainly based on traditional CMOS circuits. With the development of intelligent sensing, a large amount of environmental information needs to be collected and converted into digital signals that can be processed simultaneously, which inevitably leads to more complex circuits. CMOS circuits are complex in structure, have poor miniaturization, and are not conducive to high-density integration; moreover, CMOS-based hardware requires a large number of operations when processing complex multimodal signals, resulting in very high power consumption; in addition, in many working scenarios, the need for a large number of data transfers between sensing, storage, and processing limits the system response time. Summary of the Invention

[0004] In view of the above-mentioned defects or improvement needs of the existing technology, the present invention provides a memristor-based sensing neuron circuit and its application to solve the technical problems of poor integration and high power consumption of existing neuromorphic sensing processing hardware technology.

[0005] To achieve the above objectives, in a first aspect, the present invention provides a memristor-based sensory neuron circuit, comprising: a sensory neuron array; wherein each sensory neuron in the array converts the external environmental information it senses into a pulse signal for output in parallel;

[0006] Each sensing neuron in the array includes a resistance sensor and a threshold-switching memristor whose positive terminal is connected to the resistance sensor; the other end of the sensor on the same row is connected to the same word line of the array for power supply; the negative terminal of the memristor on the same column is connected to the same bit line of the array and grounded; the initial state of the memristor is a high-resistance state.

[0007] In the process of sensing information about the external environment, the resistance of the resistive sensor changes with the changes in the external environment, which in turn causes the voltage across the memristor connected to it to change. When the voltage across the memristor reaches its threshold voltage, the memristor undergoes a threshold transition to a low-resistance state and begins to discharge, outputting a pulse through the positive terminal until the voltage across the memristor is lower than its holding voltage, at which point the memristor returns to a high-resistance state.

[0008] More preferably, the aforementioned resistance sensor is a pressure sensor used to sense changes in the pressure of the external environment. Its resistance decreases as the external environmental pressure increases, thereby increasing the pulse frequency output by the sensing neuron.

[0009] More preferably, the material of the memristor is a phase change material that undergoes a metal-insulator transition, or a photosensitive resistive switching oxide material.

[0010] More preferably, when the material of the memristor is a phase change material of metal-insulator transition or a photosensitive resistive switching oxide material, the memristor is also used to sense the temperature change of the external environment. Its threshold voltage decreases as the temperature of the external environment increases, thereby causing the pulse amplitude of the sensing neuron to decrease and the frequency to increase.

[0011] More preferably, when the material of the memristor is a photosensitive resistive switching oxide material, the memristor is also used to sense changes in the intensity of ambient light. Its resistance decreases as the intensity of ambient light increases, thereby increasing the pulse frequency of the sensory neuron while keeping the amplitude constant.

[0012] In a second aspect, the present invention provides a signal recognition circuit, comprising: a sensory neuron circuit and a memristor neural network circuit; the sensory neuron circuit is the sensory neuron circuit provided in the first aspect of the present invention.

[0013] The memristor neural network circuit includes: a memristor input neuron circuit, a memristor synapse array, and a memristor output neuron circuit connected in sequence.

[0014] The memristor input neuron circuit includes multiple memristor input neurons, and each memristor input neuron is connected one-to-one with the positive terminal of the memristor of each sensor neuron in the sensor neuron array.

[0015] The memristor synapses in the memristor synapse array are non-volatile memristors;

[0016] The memristor output neuron circuit includes multiple memristor output neurons;

[0017] In training mode, multiple sets of external environmental information are preset. The sensory neuron circuit is used to acquire the pulse signal under each set of external environmental information and input it into the memristor neural network circuit for training.

[0018] In recognition mode, the sensory neuron circuit is used to convert the external environmental information it perceives into pulse signals and output them to the memristor neural network circuit for the recognition of the external environmental information.

[0019] More preferably, when the resistance sensor in the sensing neuron circuit is a pressure sensor, the aforementioned external environmental information includes external pressure information.

[0020] More preferably, when the material of the memristor is a phase change material that undergoes a metal-insulator transition or a photosensitive resistive switching oxide material, the aforementioned external environmental information also includes external temperature information.

[0021] More preferably, when the material of the memristor is a photosensitive resistive switching oxide material, the aforementioned external environmental information also includes external light signal information.

[0022] Thirdly, the present invention provides an electronic chip, including the sensory neuron circuit provided in the first aspect of the present invention, or the signal recognition circuit provided in the second aspect of the present invention.

[0023] In summary, the above-described technical solutions conceived in this invention can achieve the following beneficial effects:

[0024] 1. This invention provides a memristor-based sensory neuron circuit. An integrated sensory neuron array is constructed using word lines and bit lines. Each sensory neuron in the array converts its sensed external environmental information into pulse signals for output in parallel. Each sensory neuron in the array includes a series-connected resistive sensor and a threshold-changing memristor. The resistive sensor senses the external environmental signal, and its resistance changes with the sensed signal, thus acting as an adjustable resistor. This allows the memristor to operate in a local active region for normal pulse firing, serving both to sense environmental information and to divide voltage. It is an adjustable resistor with sensing capabilities. Furthermore, the resistive sensor reduces encoding errors in the sensory encoding circuit caused by the differences between memristor components when integrating multiple sensor circuits. In addition, the parasitic capacitance of the memristor itself acts as the necessary capacitance in the memristor neuron circuit, simulating changes in the neuron's membrane potential through integration, eliminating the need for additional external capacitors. This design simplifies the circuit structure, making the sensory neuron circuit more compact and significantly increasing integration density. The reduced area and hardware savings also lower energy consumption and improve energy efficiency.

[0025] 2. Furthermore, the memristor in the sensing neuron circuit provided by this invention is made of a phase-change material that transitions from metal to insulator or a photosensitive resistive switching oxide material. This allows for the processing of signals from external sensors and the sensing of ambient light, electricity, and heat, enabling the coupling of the sensed information with the sensor's modal information, thus achieving multimodal sensing. In addition, memristors based on these materials have fast switching speeds, low power consumption, and can convert external information sensed by the sensor into pulse signals without an analog-to-digital converter. This helps reduce the power consumption of the multimodal sensing neuron array and improve computational accuracy.

[0026] 3. The present invention also provides a signal recognition circuit, including the perceptual neuron circuit and the memristor neural network circuit provided in the first aspect of the present invention; the perceptual neuron circuit is the perceptual neuron circuit provided in the first aspect of the present invention; wherein, the memristor input neuron circuit includes multiple memristor input neurons, each memristor input neuron is connected one-to-one with the positive terminal of the memristor of each perceptual neuron in the perceptual neuron array, which can simultaneously and in parallel sense and process information of multiple signal points, and the signal output by the cross array can be directly transmitted to the memristor neural network circuit for recognition calculation. This parallel sensing and calculation mode greatly improves the efficiency of processing large-scale information, further improves energy efficiency, and reduces hardware area. Attached Figure Description

[0027] Figure 1 A schematic diagram of a memristor-based sensory neuron circuit structure is provided in an embodiment of the present invention;

[0028] Figure 2 A characteristic curve of a memristor IV for pulse coding provided in an embodiment of the present invention;

[0029] Figure 3 A schematic diagram of the specific structure of a sensory neuron provided in an embodiment of the present invention;

[0030] Figure 4 The diagram shows the pulse firing characteristics of the sensory neuron under different resistance values ​​R provided in the embodiments of the present invention; wherein, (a)-(e) are schematic diagrams of the pulse signals emitted by the sensory neuron under different resistance values ​​R; and (f) is a statistical diagram of the frequency of the pulse signals emitted by the sensory neuron under different resistance values ​​R.

[0031] Figure 5 The following are schematic diagrams of the IV characteristics of a memristor device at different temperatures and the circuit structure of a sensory neuron provided in the embodiments of the present invention; wherein, (a) is a schematic diagram of the IV characteristics of a memristor device at different temperatures; and (b) is a schematic diagram of the specific circuit structure of a sensory neuron unit.

[0032] Figure 6 These are output pulse characteristic diagrams of the sensing neuron under the same pressure but different temperatures in an embodiment of the present invention; wherein, (a)-(d) are schematic diagrams of the output pulse signals of the sensing neuron under the same pressure but different temperatures;

[0033] Figure 7 These are characteristic graphs of the pulse amplitude and pulse frequency output by the sensing neuron provided in the embodiments of the present invention as a function of temperature and pressure; wherein, (a) is a characteristic graph of the pulse amplitude output by the sensing neuron provided in the embodiments of the present invention as a function of temperature and pressure; and (b) is a characteristic graph of the pulse frequency output by the sensing neuron provided in the embodiments of the present invention as a function of temperature and pressure.

[0034] Figure 8 This is a diagram illustrating the application of the signal recognition circuit provided in this embodiment of the invention to the sensing of multimodal signals of temperature and pressure on the palm.

[0035] Figure 9 This is a diagram illustrating the process and results of the memristor neural network circuit after perception performing neuromorphic calculations on pulse signals, as provided in the embodiments of the present invention. (a) is a schematic diagram of the process of the memristor neural network circuit after perception performing neuromorphic calculations on pulse signals; (b) is a schematic diagram of the structure of the memristor neural network circuit; and (c) is a curve showing the change in recognition accuracy of the signal recognition circuit provided in the embodiments of the present invention with respect to the period. Detailed Implementation

[0036] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.

[0037] To achieve the above objectives, firstly, such as Figure 1 As shown, the present invention provides a memristor-based sensory neuron circuit, comprising: a sensory neuron array; each sensory neuron in the array converts the external environmental information it senses into a pulse signal for output in parallel;

[0038] Each sensing neuron in the array includes a resistance sensor and a threshold-switching memristor whose positive terminal is connected to the resistance sensor; the other end of the sensor on the same row is connected to the same word line of the array for power supply; the negative terminal of the memristor on the same column is connected to the same bit line of the array and grounded; the initial state of the memristor is a high-resistance state.

[0039] In the process of sensing information about the external environment, the resistance of the resistive sensor changes with the changes in the external environment, which in turn causes the voltage across the memristor connected to it to change. When the voltage across the memristor reaches its threshold voltage, the memristor undergoes a threshold transition to a low-resistance state and begins to discharge, outputting a pulse through the positive terminal until the voltage across the memristor is lower than its holding voltage, at which point the memristor returns to a high-resistance state.

[0040] The aforementioned sensory neuron array is an integration of a compact pulse-coded circuit, and its processed output signal V out It is a pulse signal.

[0041] The aforementioned sensory neuron array is a multimodal sensory neuron cross array, comprising memristors connected according to a preset number of rows and columns, and sensors connected to each row and column of memristors. The sensors are used to sense changes in the external multimodal environment and convert environmental information into electrical changes. The sensors can be changed according to the required sensing information, i.e., different sensors can be used. Each unit (sensing neuron) at a specific row and column position in the sensory neuron array is a 1T1S neuron unit composed of a sensor and a memristor connected in series. The integrated sensory neuron array connects the memristors in each column of the array together via bit lines and connects the sensors in each row via word lines; the word lines and bit lines constitute the connection part of the row and column devices of the sensory neuron array. The sensory neuron array provided by this invention can simultaneously and in parallel sense and process information from multiple signal points.

[0042] Each neuron unit in the sensory neuron array consists of a memristor and a sensor; one end of the sensor is connected to the power supply V. in One end is connected to the positive terminal of the memristor; the positive terminal of the memristor is connected to the sensor, and the negative terminal is grounded. Memristor neuron circuits often require a capacitor as an integrator to change the membrane potential. This is typically done with an external capacitor, one end connected to the positive terminal of the memristor and the other end grounded. However, this invention uses the parasitic capacitance C of the memristor as the capacitor in the sensing neuron circuit of the array. Compact neuron circuits based on parasitic capacitance can improve the integration density of the array and reduce energy consumption. For example... Figure 2 The figure shown is an IV characteristic curve of the memristor used for pulse coding in this invention. This memristor device exhibits good cycle uniformity. From Figure 2 As can be seen, the memristor used should have the following characteristics:

[0043] (1) Possesses a threshold voltage V th and holding voltage V h ;

[0044] (2) It has volatile threshold change characteristics.

[0045] Specifically, as the input voltage gradually increases, exceeding the device's threshold voltage V... th When the voltage drops below the holding voltage V, the device switches to a low-impedance state. At this point, the input voltage is gradually reduced until the voltage across the device falls below the holding voltage V. hold When this happens, the device will automatically return to a high-resistivity state.

[0046] like Figure 3 The diagram shown is a schematic representation of the specific structure of the sensory neuron provided by this invention (where C is the parasitic resistance of the memristor itself, located within the memristor, and is only shown for illustration). This circuit consists of the aforementioned resistive sensor and a memristor with threshold switching characteristics connected in series. The resistive sensor is positioned above the threshold switching memristor and connected to the positive terminal of the memristor. The positive voltage signal V of the memristor... out As the output pulse signal.

[0047] A resistive sensor detects environmental signals, and its resistance changes with the detected signals. Therefore, it can act as an adjustable resistor, allowing the memristor to operate in a localized active region and emit pulses normally, thus sensing environmental information and performing voltage division. It is an adjustable resistor with sensing capabilities. Furthermore, when integrating multiple sensing circuits, the resistive sensor can reduce encoding errors in the sensing encoding circuit caused by the differences between memristor components. In addition, the parasitic capacitance of the memristor itself acts as the necessary capacitance in the memristor neuron circuit, performing an integration function to simulate changes in the neuron's membrane potential, eliminating the need for an external capacitor. This design simplifies the circuit structure, making the neuron circuit more compact and increasing integration density. The reduced area and hardware savings also improve energy efficiency.

[0048] Specifically, the input voltage V in First, the parasitic capacitance C is charged through a charging circuit consisting of the resistive sensor and the parasitic capacitance C of the memristor itself. As the capacitor voltage increases, the voltage across the memristor also increases. When the threshold voltage V of the memristor is reached... th When the memristor undergoes a threshold transition and switches to a low-resistance state, the current flowing through it surges. Then, the parasitic capacitance C rapidly discharges through the circuit formed by the memristor and the capacitor, causing the voltage across the memristor to continuously decrease until it falls below the holding voltage V. h The memristor returns to a high-resistance state, at which point the current through the memristor returns to near zero, the parasitic capacitance C discharges completely, and the input voltage V... in Under the stimulation of [something], another cycle of charging and discharging begins, thus generating a series of voltage pulses V at the output of the neuron circuit. out .

[0049] like Figure 4The diagram shows the pulse firing characteristics of the memristor neuron under different resistance sensor values ​​R. It can be seen that, with other conditions remaining constant, the pulse firing frequency of the memristor neuron decreases significantly as the adjustable resistance R increases. Increasing R directly leads to a decrease in charging current, resulting in a longer time required to charge to the same threshold voltage and accumulate the same charge, thus increasing the integration time and causing a decrease in the neuron's firing frequency. This result demonstrates the influence of circuit parameters on the pulse firing characteristics of the memristor neuron, and also shows the influence of the sensor in each memristor neuron in the sensor array on the pulse frequency encoding.

[0050] In summary, this invention connects the sensors in the neuron array according to a preset number of rows and columns, configures the threshold switching memristor for each row of sensors and connects them to form a cross array structure, which saves hardware usage, has good miniaturization, simple circuit structure, is conducive to high-density and large-scale array integration, and greatly improves integrability.

[0051] In one optional implementation, the aforementioned resistance sensor is a pressure sensor used to sense changes in the pressure of the external environment. Its resistance decreases as the external environmental pressure increases, thereby increasing the pulse frequency output by the sensing neuron while keeping the amplitude essentially unchanged.

[0052] In one optional implementation, the aforementioned resistive sensor is a photosensitive sensor used to sense changes in the intensity of light in the external environment. Its resistance decreases as the intensity of light increases, thereby increasing the pulse frequency output by the sensing neuron while keeping the amplitude essentially unchanged.

[0053] In one optional implementation, the resistance sensor is a curvature sensor used to sense the curvature change of the object being measured. When the surface is bent, ductile deformation occurs, causing a change in the resistance of the diaphragm, and the curvature change of the object being measured is then converted into an electrical signal output.

[0054] In one alternative embodiment, the material of the memristor is a phase change material that undergoes a metal-insulator transition (e.g., VO). X NbO XDevices constructed using photosensitive resistive switching oxide materials (e.g., ITO, IGZO, perovskite materials, etc.), metal-insulator phase change materials, or photosensitive resistive switching oxide materials as functional layers exhibit resistive switching characteristics that change with ambient temperature. In this case, the memristor also senses changes in ambient temperature, with its threshold voltage decreasing as the ambient temperature increases, thereby reducing the amplitude and increasing the frequency of the pulses output by the sensing neuron. If a photosensitive oxide material is used, in addition to sensing temperature, it can also respond to ambient light signals. Devices constructed using photosensitive resistive switching oxide materials as functional layers exhibit resistive switching characteristics that change with ambient light intensity. In this case, when the memristor material is a photosensitive resistive switching oxide material, it also senses changes in ambient light intensity, with its resistance decreasing as the ambient light intensity increases, thereby increasing the frequency and keeping the amplitude of the pulses output by the sensing neuron essentially constant. Multimodal sensing can be achieved through the combination of the memristor and the sensor. In addition, the memristor materials mentioned above have fast switching speeds, low device power consumption, and can convert external information sensed by the sensor into pulse signals without the need for an analog-to-digital converter. This helps to reduce the power consumption of the sensory neuron array and improve computational accuracy.

[0055] To further illustrate the multimodal sensing process described above, a specific embodiment will be described in detail below:

[0056] In this embodiment, a pressure sensor is selected, and the memristor is selected with resistive switching characteristics that change with ambient temperature, and the resistive switching layer material is VO. x or NbO x Phase change materials that undergo a metal-insulator transition, where the phase change is primarily a thermally induced phase change. Figure 5 These are the IV characteristic curves of the memristor device at different temperatures and the schematic diagram of the sensory neuron circuit structure in this invention; wherein, Figure (a) is the IV characteristic curve of the memristor device at different temperatures; Figure (b) is the schematic diagram of the specific circuit structure of the sensory neuron unit. From Figure 5 As can be seen in Figure (a), the threshold voltage V of the device decreases with increasing temperature. th and holding voltage V h Both decrease, and V th V h It exhibits stronger temperature dependence, with the decrease becoming more pronounced as temperature increases. Therefore, the hysteresis window of the memristor's IV characteristic curve also decreases continuously with increasing temperature. The resistive switching process of the memristor device in this embodiment is a complex electrothermal coupling process. The threshold transition phenomenon is related to the positive feedback of current and temperature within the device. Therefore, when the ambient temperature rises, the resistive switching dynamics of the device accelerates, and the thermal feedback effect intensifies, which reduces the voltage V required for the threshold transition to occur. thBesides this, Figure 5 Figure (b) also shows the specific circuit structure of the sensory neuron unit in this embodiment. Specifically, the circuit consists of the pressure sensor described above and a temperature-sensing memristor with threshold switching characteristics connected in series, wherein the positive voltage signal V of the memristor is... out As the output pulse signal, the pressure sensor acts as an adjustable resistor, sensing pressure and dividing it; while the temperature-sensing memristor senses temperature, processes pressure signals, and couples multimodal information. In addition, the parasitic capacitance of the memristor itself acts as a necessary capacitance in the memristor neuron circuit, playing an integral role in simulating changes in the neuron's membrane potential.

[0057] The memristor in this invention serves as the core processing unit, which can not only process the pressure signals from external sensors, but also sense ambient temperature information, thereby enabling the coupling of its temperature information with other modal information. Figure 6 This diagram shows the output pulse characteristics of the sensing neuron in this embodiment under the same pressure but different temperatures. It can be seen that, under the same pressure, as the temperature increases, the amplitude of the output pulse of the multimodal sensing neuron circuit decreases while the frequency increases. This is because the device threshold voltage V increases with temperature. th A decrease in temperature leads to a faster integration process in the multimodal sensing neuron circuit, thereby increasing the frequency of neuron firing pulses. Simultaneously, it reduces the amplitude of the output pulses. Therefore, the amplitude and firing frequency of the output pulses encode ambient temperature information, which is crucial for achieving temperature sensing in this invention. To more clearly understand the impact of temperature and pressure on the output pulses of neuron units in the multimodal sensing neuron array of this invention, Figure 7 The diagram further illustrates the variation characteristics of the pulse amplitude and frequency output by the sensing neuron in this embodiment with temperature and pressure. It can be seen that at the same temperature, as the external pressure sensed by the sensor increases, the amplitude of the pulse output by the sensing neuron remains essentially constant, while the firing frequency increases significantly. When the external pressure increases, the resistance R of the pressure sensor decreases, leading to an increase in charging current. This results in less time required for the capacitor to accumulate the same charge. With a fixed device threshold voltage, the amplitude of the voltage pulse output by the sensing neuron remains essentially constant, while the firing frequency increases significantly. Therefore, the firing frequency of the pulse can encode the external pressure information sensed by the pressure sensor. In summary, the pulse amplitude only decreases with increasing temperature and remains essentially constant with increasing pressure; therefore, the pulse amplitude cannot encode the multimodal environment (pressure and temperature) in this embodiment. However, an increase in either temperature or pressure will cause an increase in the pulse firing frequency. Therefore, since the memristor can encode complex analog signals (temperature and pressure) in the environment into pulse signals of different frequencies, this invention can achieve pulse encoding of external multimodal signals.

[0058] Based on this, the sensory neuron array in this embodiment can perform pulse encoding of external multimodal signals, encoding complex analog signals in the environment into pulse signals of different frequencies, thereby achieving multimodal fusion sensing and computation. Specifically, the sensors in the array can change according to the required sensing information, i.e., different sensors can be used. The memristor, as the core processing unit, can not only process signals from external sensors, but also sense ambient light / electricity / heat information, so as to achieve coupling between its sensing information and the modal information of the sensors. Compared with the prior art, this invention makes full use of the sensing encoding and compact structure characteristics of memristor neural circuits to design a multimodal sensing array, effectively solving the problems of complex circuit integration, high latency, and high energy consumption in traditional sensing processing integrated systems, saving hardware usage, and having the advantages of parallel processing, high density, small area, and low energy consumption, which can be applied to edge intelligent sensing computing scenarios.

[0059] Secondly, the present invention provides a signal recognition circuit, including: a sensory neuron circuit and a memristor neural network circuit; the sensory neuron circuit is the sensory neuron circuit provided in the first aspect of the present invention; the related technical solutions are the same as the sensory neuron circuit provided in the first aspect of the present invention, and will not be described in detail here;

[0060] The memristor neural network circuit includes: a memristor input neuron circuit, a memristor synapse array, and a memristor output neuron circuit connected in sequence.

[0061] The memristor input neuron circuit includes multiple memristor input neurons, all of which are non-volatile memristors; each memristor input neuron is connected one-to-one with the positive terminal of the memristor of each sensor neuron in the sensory neuron array.

[0062] The memristor synapses in the memristor synapse array are non-volatile memristors;

[0063] The memristor output neuron circuit includes multiple memristor output neurons, all of which are non-volatile memristors;

[0064] In training mode, multiple sets of external environmental information (such as setting pressure and / or temperature) are preset. The sensory neuron circuit is used to acquire the pulse signal under each set of external environmental information and input it into the memristor neural network circuit for training (the training label is the corresponding external environmental information).

[0065] In recognition mode, the sensory neuron circuit is used to convert the external environmental information it senses into pulse signals and output them to the memristor neural network circuit to identify and classify the external environmental information.

[0066] It should be noted that the memristors used in memristor neural network circuits are all non-volatile devices with multi-resistivity characteristics.

[0067] In one alternative implementation, when the resistance sensor in the sensing neuron circuit is a pressure sensor, the aforementioned external environment information includes external pressure information.

[0068] In one alternative implementation, when the material of the memristor is a phase change material that transitions from metal to insulator or a photosensitive resistive switching oxide material, the aforementioned external environmental information also includes external temperature information.

[0069] If a photosensitive oxide material is used, in addition to sensing temperature, it can also respond to light signals in the environment. In this case, the aforementioned external environmental information also includes external light signal information.

[0070] In one specific embodiment, such as Figure 8 The diagram illustrates the application of the signal recognition circuit provided by this invention to the sensing of multimodal signals of temperature and pressure on the palm. This embodiment applies a sensory neuron array to the sensing of multimodal signals of temperature and pressure on the palm and uses a memristor neural network circuit to calculate and recognize them. By encoding the multimodal signals (temperature and pressure signals) simultaneously present on the palm into different voltage signals as inputs to the sensory neuron array, the array can sense and calculate these input signals and encode them into pulse signals of different frequencies for output. These output pulse signals are then used as inputs to the memristor neural network circuit for neuromorphic calculation and recognition. The color at a certain point on the palm represents the normalized frequency intensity corresponding to the temperature and pressure coupling condition, thus reflecting the pressure and temperature magnitude at that point on the palm. This completes the sensing and encoding function.

[0071] Then comes the neuromorphic computation part after perception, which consists of memristor neuron circuits (memristor input neuron circuits and memristor output neuron circuits) and memristor synaptic arrays. Each memristor input neuron in the memristor input neuron circuit is connected to each row and column in the sensory neuron array. Figure 9 This diagram illustrates the process and results of the memristor neural network circuit in this invention performing neuromorphic calculations on pulse signals after perception. This part processes pulse signals of different frequencies output by the sensory neuron array, which are related to multimodal information on the palm, and performs identification and classification. A memristor neuromorphic calculation system uses a memristor-based neuromorphic chip, which can receive processed pulse signals and perform calculations and identifications through a spiking neural network. The memristor synaptic array in the memristor neuromorphic chip contains memristor devices, which are non-volatile devices with multi-resistive-state characteristics. As can be seen from the results diagram, the memristor neural network in this embodiment achieved a recognition accuracy of 97% after multiple training cycles, demonstrating a high-precision recognition effect on multimodal information on the palm.

[0072] In summary, this invention proposes a memristor-based sensing neuron circuit and a corresponding signal recognition circuit, which can encode the modal signals in the sensed environment into pulse signals of different frequencies and input them into a memristor neural network for recognition and calculation. Compared with traditional CMOS sensing processing hardware, this invention cleverly utilizes the threshold switching characteristics and temperature-dependent resistive switching characteristics of volatile memristors to achieve multimodal information processing on the palm of the hand. It eliminates the need for high-overhead analog-to-digital converters and complex control circuits, greatly simplifying circuit design, reducing energy consumption, and facilitating high-density large-scale integration. It has excellent application prospects and promotes the development of neuromorphic sensing and computing. Furthermore, compared with existing single-memristor neuron circuits, the multimodal sensing neuron array in this invention has multiple sensing points, enabling simultaneous sensing of large-scale information, further improving energy efficiency and reducing hardware area.

[0073] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A memristor-based sensory neuron circuit, characterized in that, include: A sensory neuron array; each sensory neuron in the array converts the external environmental information it senses into pulse signals for output in parallel; Each sensory neuron in the array includes a resistance sensor and a threshold-switching memristor whose positive terminal is connected to one end of the resistance sensor; the other end of the resistance sensor in the same row is connected to the same word line of the array for power supply; the negative terminal of the memristor in the same column is connected to the same bit line of the array and grounded; the initial state of the memristor is a high-resistance state. In the process of sensing information about the external environment, the resistance of the resistive sensor changes with the changes in the external environment, which in turn causes the voltage across the memristor connected to it to change. When the voltage across the memristor reaches its threshold voltage, the memristor undergoes a threshold transition to a low-resistance state and begins to discharge, outputting a pulse through the positive terminal until the voltage across the memristor is lower than its holding voltage, at which point the memristor returns to a high-resistance state.

2. The sensory neuron circuit according to claim 1, characterized in that, The resistance sensor is a pressure sensor used to sense changes in the pressure of the external environment. Its resistance decreases as the external environmental pressure increases, thereby increasing the pulse frequency output by the sensing neuron.

3. The sensory neuron circuit according to claim 1 or 2, characterized in that, The memristor is made of a phase change material that transitions from metal to insulator or a photosensitive resistive oxide material.

4. The sensory neuron circuit according to claim 3, characterized in that, When the material of the memristor is a phase change material of metal-insulator transition or a photosensitive resistive switching oxide material, the memristor is also used to sense the temperature change of the external environment. Its threshold voltage decreases as the temperature of the external environment increases, thereby causing the pulse amplitude of the sensing neuron to decrease and the frequency to increase.

5. The sensory neuron circuit according to claim 3, characterized in that, When the material of the memristor is a photosensitive resistive switching oxide material, the memristor is also used to sense changes in the intensity of ambient light. Its resistance decreases as the intensity of ambient light increases, thereby increasing the pulse frequency of the sensing neuron while keeping the amplitude constant.

6. A signal recognition circuit, characterized in that, include: Sensory neuron circuits and memristor neural network circuits; The sensory neuron circuit is the sensory neuron circuit according to any one of claims 1-5; The memristor neural network circuit includes: a memristor input neuron circuit, a memristor synapse array, and a memristor output neuron circuit connected in sequence; The memristor input neuron circuit includes multiple memristor input neurons, and each memristor input neuron is connected to the positive terminal of the memristor of each sensor neuron in the sensory neuron array. The memristor synapses in the memristor synapse array are non-volatile memristors; The memristor output neuron circuit includes multiple memristor output neurons; In training mode, multiple sets of external environmental information are preset. The sensory neuron circuit is used to acquire pulse signals under each set of external environmental information and input them into the memristor neural network circuit for training. In recognition mode, the sensory neuron circuit is used to convert the external environmental information it senses into pulse signals and output them to the memristor neural network circuit to recognize the external environmental information.

7. The signal recognition circuit according to claim 6, characterized in that, When the resistance sensor in the sensory neuron circuit is a pressure sensor, the external environmental information includes external pressure information.

8. The signal recognition circuit according to claim 6 or 7, characterized in that, When the material of the memristor is a phase change material that transitions from metal to insulator or a photosensitive resistive switching oxide material, the external environmental information also includes external temperature information.

9. The signal recognition circuit according to claim 6 or 7, characterized in that, When the material of the memristor is a photosensitive resistive switching oxide material, the external environment information also includes external light signal information.

10. An electronic chip, characterized in that, It includes the sensory neuron circuit according to any one of claims 1-5, or the signal recognition circuit according to any one of claims 6-9.

Images