An artificial neuron circuit based on volatile threshold transition devices

Abstract

The application provides an artificial neuron circuit based on a volatile threshold transition device, and belongs to the field of artificial neural morphology; wherein, a membrane potential generation circuit is used for generating different membrane potentials under different input currents and different external reset voltages; a slow variable generation circuit is used for generating a slow variable; a membrane potential reset circuit comprises a volatile threshold transition device and an external reset voltage excitation source; the external reset voltage excitation source is used for providing different external reset voltages, so that the membrane potential and the slow variable change, and the biological neuron discharge behavior is simulated; and the volatile threshold transition device is used for realizing the membrane potential reset function. Compared with a traditional neuron based on a CMOS circuit, the neuron circuit provided in the application has the characteristics of simple structure, strong flexibility and rich functions, and the number of transistors required is greatly reduced, and the hardware and power consumption overheads are reduced, which is beneficial to large-scale integration in a hardware pulse neural network.

Description

Technical Field

[0001] This application belongs to the field of artificial neuromorphology, and more specifically, relates to an artificial neuron circuit based on a volatile threshold switching device. Background Technology

[0002] Biologically inspired artificial neuromorphic computing architectures are considered an effective means to overcome the von Neumann bottleneck. Through the cross-interconnection of artificial neurons and synapses, they can achieve brain-like computing functions with high energy efficiency, low latency and high parallelism, and have great application prospects in the field of artificial intelligence.

[0003] Biological neurons are characterized by high nonlinearity and rich dynamics. Firing behavior is one of the main electrical activities of biological neurons, playing a crucial role in information transmission, processing, and encoding. Known firing behaviors of biological neurons include regular firing (RS), intrinsically bursting (IB), chatting (CH), and fast firing (FS).

[0004] In recent years, researchers have constructed various analog and digital circuits to simulate the high-order output characteristics of biological neurons. Among them, artificial neurons represented by the Izhikevich model are built based on CMOS circuits. This model is very suitable for simulating large-scale networks, but such neuron circuits require at least dozens of transistors, have complex circuit structures, and have high power consumption and area overhead, which is not conducive to large-scale integration.

[0005] Volatile threshold switching devices (VTs) have been widely used in the construction of artificial neuromorphic hardware due to their simple structure and rich dynamic characteristics. Currently, neuronal circuits based on VTs have been reported; however, most circuits are simplified models of Leaky Integrate-and-Fire (LIF) neurons, exhibiting a single firing mode and failing to simulate the diverse firing behaviors of biological neurons, thus hindering the construction of biomimetic high-order brain-like systems.

[0006] Therefore, there is an urgent need for a simple artificial neuron circuit to realize the firing behavior of various biological neurons. Summary of the Invention

[0007] To address the shortcomings of existing technologies, this application aims to provide an artificial neuron circuit based on a volatile threshold switching device. This addresses the problem that current neuron circuits based on volatile threshold switching devices are simplified models of leakage current integral ignition neurons, which have a single firing mode and cannot simulate the diverse firing behaviors of biological neurons, thus making it difficult to construct biomimetic high-order brain-like systems.

[0008] To achieve the above objectives, in a first aspect, this application provides an artificial neuron circuit based on a volatile threshold switching device, comprising: a membrane potential generation circuit, a slow variable generation circuit, and a membrane potential reset circuit;

[0009] The membrane potential generation circuit is used to generate different membrane potentials under different input currents and different external reset voltages; the membrane potential generation circuit is connected to the slow variable generation circuit; the slow variable generation circuit is used to generate slow variables.

[0010] The membrane potential reset circuit is connected to the membrane potential generation circuit and includes a volatile threshold switching device and an external reset voltage excitation source. The volatile threshold switching device is connected to the external reset voltage excitation source. The external reset voltage excitation source is used to provide different external reset voltages to change the membrane potential and slow variables, simulating the firing behavior of biological neurons. The volatile threshold switching device is used to realize the membrane potential reset function.

[0011] The external reset voltage is less than the holding voltage of the volatile threshold switching device.

[0012] More preferably, the membrane potential generation circuit includes a first capacitor, a first NMOS transistor, a second PMOS transistor, a third PMOS transistor, and a fourth NMOS transistor; one end of the first capacitor is connected to the gate of the first NMOS transistor, and the other end is connected to the source of the first NMOS transistor and grounded; the first capacitor is connected to a volatile threshold switching device; the first capacitor is used to drive the first NMOS transistor to generate current using voltage; the second PMOS transistor and the third PMOS transistor form a current mirror circuit, used to mirror the current of the first NMOS transistor to the third PMOS transistor proportionally; the current difference between the third PMOS transistor and the fourth NMOS transistor flows into the first capacitor, used to form a positive feedback loop, generating different membrane potentials under different input currents and different external reset voltages.

[0013] More preferably, the slow variable generation circuit includes a second capacitor, a fifth PMOS transistor, and a sixth NMOS transistor; the fifth PMOS transistor and the second PMOS transistor constitute another set of current mirror circuits, used to mirror the current of the first NMOS transistor to the fifth PMOS transistor proportionally; one end of the second capacitor is connected to the gate of the fourth NMOS transistor, and the other end is connected to the source of the fourth NMOS transistor and grounded; the second capacitor is used to drive the fourth NMOS transistor to generate current using voltage; the gate and source of the sixth NMOS transistor are shorted, so that the current difference between the fifth PMOS transistor and the sixth NMOS transistor flows into the second capacitor, generating a slow variable.

[0014] More preferably, when the difference between the membrane potential and the external reset voltage is greater than the threshold voltage of the volatile threshold switching device, the volatile threshold switching device changes from a high-resistance state to a low-resistance state, and the first capacitor is used to discharge through the volatile threshold switching device; when the difference between the membrane potential and the external reset voltage is less than the holding voltage of the volatile threshold switching device, the volatile threshold switching device changes from a low-resistance state to a high-resistance state, which is used to reset the membrane potential.

[0015] More preferably, the volatile threshold switching device is a threshold switching device based on metal-insulator transition, based on metal conductive filament diffusion, or based on chalcogenides;

[0016] Threshold switching devices based on metal-insulator transitions employ NbO x or VO x Mott insulator material; threshold switching devices based on diffusion of metal conductive filaments use Ag or Cu active metal materials to form conductive filaments; threshold switching devices based on chalcogenides use tellurium-based, selenium-based, or sulfide-based OTS materials doped with single or multiple elements; where x > 0.

[0017] More preferably, the first capacitor and the second capacitor are fixed capacitors or variable capacitors, with values ​​ranging from 1fF to 1μF.

[0018] More preferably, the firing behavior of biological neurons includes regular firing, initiation cluster firing, cluster firing, and rapid firing.

[0019] More preferably, the state equation of the artificial neuron circuit is:

[0020]

[0021] Where V and U are the membrane potential and the slow variable, respectively; I in I3 is the input current; I5 and I4 are the currents of the third and fifth PMOS transistors, respectively; I5 and I6 are the currents of the fourth and sixth NMOS transistors, respectively; C v C is the capacitance value of the first capacitor; u This is the capacitance value of the second capacitor.

[0022] Secondly, this application provides a method for firing artificial neurons based on a volatile threshold switching device, specifically including the following steps:

[0023] Step S1: Under the application of external reset voltage and input current, the membrane potential rises rapidly through the positive feedback loop of the membrane potential generation circuit. When the difference between the membrane potential and the external reset voltage is greater than the threshold voltage of the volatile threshold switching device, the volatile threshold switching device changes from a high-resistance state to a low-resistance state, providing a low-resistance discharge path for the first capacitor, causing the membrane potential to drop rapidly. The current of the first NMOS transistor increases rapidly and then decreases. Through the current mirror circuit, the current corresponding to the fifth PMOS transistor also increases rapidly and then decreases, charging the second capacitor and increasing the slow variable.

[0024] Step S2: When the difference between the membrane voltage and the external reset voltage is less than the holding voltage of the volatile threshold switching device, the volatile threshold switching device changes from a low-resistance state to a high-resistance state; the membrane potential is continuously charged and reset, the slow variable continues to increase, and the neuron enters the initial stage of firing;

[0025] Step S3: When the slow variable increases to its maximum value, the current in the fourth NMOS transistor and the sixth NMOS transistor increases, making the rate of change of the membrane potential and the rate of change of the slow variable both less than 0, and the membrane potential and the slow variable decrease.

[0026] Step S4: During the decrease of the slow variable, the rate of change of the membrane potential is greater than 0, the rate of change of the slow variable is less than 0, the membrane potential rises, the slow variable continues to decrease, and the neuronal firing behavior enters the refractory period.

[0027] Step S5: The slow variable continues to decrease, and the rate of change of both the membrane potential and the slow variable is greater than 0, causing the membrane potential to be continuously charged and reset. Proceed to step S1 until the artificial neuron stops firing.

[0028] More preferably, the firing behavior of biological neurons includes regular firing, initiation of cluster firing, cluster firing, and rapid firing, and the external reset voltage required for rapid firing is greater than that required for cluster firing; the external reset voltage required for cluster firing is greater than that required for initiation of cluster firing; and the external reset voltage required for initiation of cluster firing is greater than that required for regular firing.

[0029] It should be noted that the rapid discharge mentioned in this application refers to a discharge mode in which the increase of the slow variable cannot make the rate of change of membrane potential less than 0 after the external reset voltage exceeds a certain voltage level, and the neuron has no refractory period during the entire discharge process.

[0030] Overall, the technical solutions conceived in this application have the following beneficial effects compared with the prior art:

[0031] This application provides an artificial neuron circuit based on a volatile threshold switching device, including a membrane potential generation circuit, a slow variable generation circuit, and a membrane potential reset circuit. Utilizing the unique electrical characteristics of the volatile threshold switching device, the membrane potential reset function is achieved, reducing the number of transistors. This circuit only requires changing an external reset voltage to simulate the firing behaviors of four known biological neurons within the same circuit. Compared to traditional CMOS-based neurons, the neuron circuit provided in this application requires a significantly reduced number of transistors, resulting in lower hardware and power consumption overhead. It features a simple structure, high flexibility, and rich functionality, which is beneficial for large-scale integration in hardware spiking neural networks. Attached Figure Description

[0032] Figure 1 This is a schematic diagram of the structure of an artificial neuron circuit based on a volatile threshold switching device provided in an embodiment of this application;

[0033] Figure 2 This is a voltage-current scan curve of a SiTe chalcogenide threshold transition device provided in an embodiment of this application;

[0034] Figure 3(a) is a time-domain waveform diagram of the regular discharge of an artificial neuron circuit based on a chalcogenide threshold switching device provided in an embodiment of this application;

[0035] Figure 3(b) is a phase diagram of the regular discharge of an artificial neuron circuit based on a chalcogenide threshold switching device provided in an embodiment of this application;

[0036] Figure 4(a) is a time-domain waveform diagram of the initiation cluster discharge of an artificial neuron circuit based on a chalcogenide threshold switching device provided in an embodiment of this application;

[0037] Figure 4(b) is a phase diagram of the initiation of cluster discharge in an artificial neuron circuit based on a chalcogenide threshold transition device provided in an embodiment of this application;

[0038] Figure 5(a) is a time-domain waveform diagram of the discharge of an artificial neuron circuit cluster based on a chalcogenide threshold switching device provided in an embodiment of this application;

[0039] Figure 5(b) is a phase diagram of the discharge of an artificial neuron circuit cluster based on a chalcogenide threshold switching device provided in an embodiment of this application;

[0040] Figure 6(a) is a time-domain waveform diagram of rapid discharge of an artificial neuron circuit based on a chalcogenide threshold switching device provided in an embodiment of this application;

[0041] Figure 6(b) is a phase diagram of rapid discharge of an artificial neuron circuit based on a chalcogenide threshold switching device provided in an embodiment of this application;

[0042] Figure 7This is a one-dimensional summary diagram of four discharge behaviors of an artificial neuron circuit based on a chalcogenide threshold switching device provided in the embodiments of this application. Detailed Implementation

[0043] The embodiments of this application are described below with reference to the accompanying drawings.

[0044] To meet the need for developing artificial neuron circuits with simple structure and high flexibility, this application provides an artificial neuron circuit based on a volatile threshold switching device, including: a membrane potential generation circuit, a slow variable generation circuit, and a membrane potential reset circuit.

[0045] The membrane potential generation circuit includes: a first capacitor C v The first NMOS transistor M1, the fourth NMOS transistor M4, the second PMOS transistor M2, and the third PMOS transistor M3; the first capacitor C v One end is connected to the gate of the first NMOS transistor M1, and the other end is connected to the source of the first NMOS transistor M1 and grounded. The first capacitor C v The voltage on the first NMOS transistor M1 drives the first NMOS transistor M1 to generate current; the second PMOS transistor M2 and the third PMOS transistor M3 form a current mirror circuit, mirroring the current of the first NMOS transistor M1 to the third PMOS transistor M3 in a certain proportion; the difference between the currents of the third PMOS transistor M3 and the fourth NMOS transistor M4 flows into the first capacitor C. v This forms a positive feedback loop, generating a membrane potential V;

[0046] The slow variable generation circuit includes: a second capacitor C u The fifth PMOS transistor M5 and the sixth NMOS transistor M6; the fifth PMOS transistor M5 and the second PMOS transistor M2 form another current mirror circuit; the current of the first NMOS transistor M1 is mirrored to the fifth PMOS transistor M5 at a certain ratio; the second capacitor C u One end is connected to the gate of the fourth NMOS transistor M4, and the other end is connected to the source of the fourth NMOS transistor M4 and grounded. The second capacitor C u The voltage on the capacitor drives the fourth NMOS transistor M4 to generate current; the gate and source terminals of the sixth NMOS transistor M6 are shorted, causing the current difference between the fifth PMOS transistor M5 and the sixth NMOS transistor M6 to flow into the second capacitor C. u This leads to the generation of a slow variable U;

[0047] The membrane potential reset circuit includes: a volatile threshold switching device and an external reset voltage V. c One end of the volatile threshold switching device is connected to the first capacitor C. vThe other end is connected to the external reset voltage V. c connect.

[0048] In some embodiments, the volatile threshold switching device can achieve an instantaneous switching from a high-resistance state to a low-resistance state; specifically, when the voltage applied to the volatile threshold switching device increases from zero to the threshold voltage V... th During the process, it is in a high-resistance state; when the voltage applied to the volatile threshold switching device exceeds the threshold voltage V... th It then enters a low-resistance state; when the voltage decreases and is less than the holding voltage V h Afterwards, the device spontaneously returns to a high-impedance state, where the threshold voltage V th Greater than the holding voltage V h .

[0049] In some embodiments, the membrane potential reset circuit utilizes the electrical characteristics of a volatile threshold switching device, when the membrane potential V is different from the external reset voltage V0. c The difference is greater than the threshold voltage V of the volatile threshold switching device. th When the volatile threshold switching device transitions from a high-resistance state to a low-resistance state, the capacitor discharges through the volatile threshold switching device; when the film potential V is equal to the external reset voltage V... c The difference is less than the holding voltage V of the volatile threshold switching device. h When the volatile threshold switching device changes from a low-resistance state to a high-resistance state, it can achieve the function of resetting the membrane potential.

[0050] In some embodiments, the external reset voltage V c The holding voltage V of the volatile threshold switching device is less than the holding voltage V. h Capacitor C v and C u It can be a fixed capacitor or a variable capacitor, with the capacitance value ranging from 1fF to 1μF.

[0051] In some embodiments, the volatile threshold switching device described above can be a threshold switching device based on metal-insulator transition, based on metal conductive filament diffusion, or based on chalcogenides; specifically, the threshold switching device based on metal-insulator transition uses NbO. x or VO x Mott insulator material; threshold switching devices based on diffusion of metal conductive filaments use Ag or Cu active metal materials to form conductive filaments; threshold switching devices based on chalcogenides use tellurium-based, selenium-based, or sulfide-based OTS materials doped with single or multiple elements; where x > 0.

[0052] In some embodiments, the artificial neuron circuit, under a certain current input, changes the external reset voltage V. cThe membrane potential generation circuit outputs different membrane potentials V, which can simulate the four known biological neuron firing behaviors; specifically, the four biological neuron firing behaviors include regular firing, initiation cluster firing, cluster firing, and rapid firing.

[0053] The above technical solution will be described in detail below with reference to specific embodiments.

[0054] This embodiment provides an artificial neuron circuit based on a volatile threshold switching device, as shown in the schematic diagram below. Figure 1 As shown; among them, the volatile threshold switching device is taken as a chalcogenide threshold switching device as an example. Figure 2 A voltage-current scan curve of a SiTe chalcogenide threshold transition device is provided, which has a threshold voltage V of 1.14V. th and a holding voltage of 0.9V V h .

[0055] Using general circuit analysis methods, the state equation of the artificial neuron circuit can be obtained from its topology in this embodiment as follows:

[0056]

[0057] Where V and U are the membrane potential and the slow variable, respectively; I in I3 and I5 are the currents of the third PMOS transistor M3 and the fifth PMOS transistor M5, respectively; I4 and I6 are the currents of the fourth NMOS transistor M4 and the sixth NMOS transistor M6; specifically, equation (1) can be obtained from the membrane potential generation circuit, and equation (2) can be obtained from the slow variable generation circuit.

[0058] Furthermore, the process of simulating the firing behavior of biological neurons is explained in detail; under a certain external reset voltage V c and input current I in Under the application of [condition], the positive feedback loop of the membrane potential generation circuit causes the membrane potential V to rise rapidly. When the membrane potential V is equal to the external reset voltage V[condition], [the membrane potential V] will rise rapidly. c The difference is greater than the threshold voltage V of the volatile threshold switching device. th At that time, the volatile threshold switching device transitions from a high-resistance state to a low-resistance state, thereby providing capacitance C. v A low-resistance discharge path is provided, causing the film potential V to drop rapidly to the set voltage c. Because the film potential V rises and then falls rapidly, the current in the first NMOS transistor M1 increases and then decreases rapidly. Through the current mirror circuit, the current I5 also increases and then decreases rapidly, supplying power to capacitor C. u When a certain charge is applied, the slow variable U increases by an incremental voltage d. The reset mechanism is as follows:

[0059] If > Vreset ,but

[0060] Among them, V reset =V th +V c c and d change with the external reset voltage V c It increases with the increase of.

[0061] When the membrane potential V is equal to the external reset voltage V c The difference is less than the holding voltage V of the volatile threshold switching device. h At this time, the volatile threshold switching device changes from a low-resistance state to a high-resistance state; after a period of time, the membrane potential V is continuously charged and reset, the slow variable U continues to increase, and the neuron enters the initial stage of firing.

[0062] From the state equation of the neuron circuit, we know that when the slow variable U increases to a certain extent, the increase in currents I4 and I6 driven by the slow variable U makes the rate of change of membrane potential V (dV / dt) and the rate of change of slow variable U (dU / dt) both less than 0, causing membrane potential V and slow variable U to decrease at a certain rate. When the slow variable U decreases to a certain extent, the rate of change of membrane potential V (dV / dt) is slightly greater than 0, but the rate of change of slow variable U (dU / dt) is still less than 0, causing membrane potential V to rise slowly and slow variable U to continue to decrease. This is the refractory period of neuronal firing behavior. Generally, the larger the voltage increment (d) of the slow variable U, the longer the refractory period of the neuron. After the neuron enters the refractory period, the membrane potential V and the slow variable U change at a certain rate. When the slow variable U decreases to a certain extent, the rates of change of both the membrane potential V (dV / dt) and the slow variable U (dU / dt) are greater than 0, causing the membrane potential V to continuously charge and reset, and the slow variable U to continuously increase. When the slow variable U increases to a certain extent, the neuron re-enters the refractory period and then cyclically produces firing behavior, entering a stable firing phase. These two phases constitute the firing behavior of the artificial neuron circuit.

[0063] When the external reset voltage V c When the voltages c and d in equation (3) change, the neuron firing behavior will also change.

[0064] When the external reset voltage V c The input current I is 0. in When the voltage is 400nA, due to the external reset voltage V cThe smaller voltage d in the reset mechanism results in a smaller slow variable U after the initial stage of neuronal firing. The rate of change of membrane potential V, dV / dt, is slightly greater than 0, but the rate of change of slow variable U, dU / dt, is less than 0, causing membrane potential V to rise slowly and slow variable U to decrease slightly. After the refractory period, neuronal firing enters the steady stage. The steady stage has more steps of decreasing and increasing slow variable U than the initial stage. Therefore, the firing frequency in the steady stage is lower than the firing frequency in the initial stage. The neuronal circuit exhibits the regular firing behavior shown in Figure 3(a), and the phase diagram is shown in Figure 3(b).

[0065] When the external reset voltage V c =300mV and input current I in When the voltage is 400nA, due to the external reset voltage V c The larger voltage d in the reset mechanism results in a larger slow variable U after the initial stage of neuronal firing. When the neuron enters the refractory period, the slow variable U decreases. Due to the change in the rate of change of membrane potential V dV / dt, the membrane potential V first decreases and then increases. When the slow variable U decreases to a certain extent, both the rate of change of membrane potential V dV / dt and the rate of change of slow variable U dU / dt are greater than 0. However, after the membrane potential V is charged and reset once, the slow variable U increases the voltage d, which just causes the neuron to re-enter the refractory period. This cycle repeats, and the neuron circuit exhibits the initial cluster firing behavior shown in Figure 4(a), and the phase diagram is shown in Figure 4(b).

[0066] When the external reset voltage V c =450mV and input current I in At 400 nA, the neuronal firing behavior is similar to that of the initial cluster firing, but the difference lies in the external reset voltage V. c The external reset voltage is larger than that of the initial cluster discharge behavior. The slow variable U after the neuron discharges through the initial stage is larger, which makes the process of the slow variable U decreasing during the refractory period of the neuron longer. After the refractory period, the membrane potential V needs to be charged and reset multiple times before it can re-enter the refractory period. The neuron circuit exhibits the cluster discharge behavior shown in Figure 5(a), and the phase diagram is shown in Figure 5(b).

[0067] When the external reset voltage V c =500mV and input current I in When the voltage is 400nA, due to the external reset voltage V c Compared to the first three discharge behaviors, the voltage c after each reset of the membrane potential V is the largest and V is the largest. resetBoth are very large. The difference between currents I3 and I4 is always greater than 0 during the discharge process. At this time, the increase of the slow variable U cannot make the rate of change of membrane potential V, dV / dt, less than 0. That is, dV / dt is greater than 0 throughout the discharge process. The neuron has no refractory period. The neuron circuit exhibits the rapid discharge behavior shown in Figure 6(a), and the phase diagram is shown in Figure 6(b).

[0068] Figure 7 This is a one-dimensional summary diagram of four discharge behaviors of an artificial neuron circuit based on a chalcogenide threshold switching device, when the external reset voltage V c Within the 0 to 180mV range, the neuronal circuit exhibits regular firing behavior; when the external reset voltage V... c Within the 180 to 320 mV range, the neuronal circuit exhibits initiation cluster firing behavior; when the external reset voltage V... c Within the 320 to 470 mV range, the neuronal circuit exhibits cluster firing behavior; when the external reset voltage V c Within the 470 to 500 mV range, the neuronal circuit exhibits rapid firing behavior;

[0069] This application provides an artificial neuron circuit based on a volatile threshold switching device, including a membrane potential generation circuit, a slow variable generation circuit, and a membrane potential reset circuit. The reset function is achieved by utilizing the unique electrical characteristics of the volatile threshold switching device, reducing the number of transistors. This circuit only requires changing an external reset voltage V. c This allows for the simulation of the firing behavior of four known biological neurons in the same circuit. Compared with traditional neurons based on CMOS circuits, the neuron circuit provided in this application requires a significantly reduced number of transistors, lowers hardware and power consumption, and features a simple structure, high flexibility, and rich functionality, which is beneficial for large-scale integration in hardware spiking neural networks.

[0070] It should be understood that expressions such as “comprising” and “may include” used in this application indicate the existence of the disclosed functions, operations or constituent elements, and do not limit one or more additional functions, operations and constituent elements.

[0071] In the description of the embodiments of this application, it should be noted that, unless otherwise explicitly specified and limited, the term "connection" should be interpreted broadly. For example, "connection" can be a detachable connection or a non-detachable connection; it can be a direct connection or an indirect connection through an intermediate medium.

[0072] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. An artificial neuron circuit based on a volatile threshold transition device, characterized by, include: Membrane potential generation circuit, slow variable generation circuit, and membrane potential reset circuit; The membrane potential generation circuit is used to generate different membrane potentials under different input currents and different external reset voltages. The slow variable generation circuit is connected to the membrane potential generation circuit, and the slow variable generation circuit is used to generate slow variables. The membrane potential reset circuit is connected to the membrane potential generation circuit, and includes a volatile threshold switching device and an external reset voltage excitation source; The volatile threshold switching device is connected to an external reset voltage excitation source; the external reset voltage excitation source is used to provide different external reset voltages to change the membrane potential and slow variables, simulating the firing behavior of biological neurons; Volatile threshold switching devices are used to realize the membrane potential reset function; Among them, the external reset voltage is less than the holding voltage of the volatile threshold switching device; When the difference between the membrane potential and the external reset voltage is greater than the threshold voltage of the volatile threshold switching device, the volatile threshold switching device changes from a high-resistance state to a low-resistance state, and the first capacitor is used to discharge through the volatile threshold switching device; when the difference between the membrane potential and the external reset voltage is less than the holding voltage of the volatile threshold switching device, the volatile threshold switching device changes from a low-resistance state to a high-resistance state, which is used to reset the membrane potential.

2. The artificial neuron circuit according to claim 1, wherein, The membrane potential generation circuit includes a first capacitor, a first NMOS transistor, a second PMOS transistor, a third PMOS transistor, and a fourth NMOS transistor. One end of the first capacitor is connected to the gate of the first NMOS transistor, and the other end is connected to the source of the first NMOS transistor and grounded. The first capacitor is connected to the volatile threshold switching device. The first capacitor is used to drive the first NMOS transistor to generate current using voltage. The second PMOS transistor and the third PMOS transistor form a current mirror circuit to mirror the current of the first NMOS transistor to the third PMOS transistor proportionally. The current difference between the third PMOS transistor and the fourth NMOS transistor flows into the first capacitor to form a positive feedback loop, generating different membrane potentials under different input currents and different external reset voltages.

3. The artificial neuron circuit according to claim 2, wherein, The slow variable generation circuit includes a second capacitor, a fifth PMOS transistor, and a sixth NMOS transistor. The fifth PMOS transistor and the second PMOS transistor form another current mirror circuit, used to mirror the current of the first NMOS transistor to the fifth PMOS transistor proportionally. One end of the second capacitor is connected to the gate of the fourth NMOS transistor, and the other end is connected to the source of the fourth NMOS transistor and grounded. The second capacitor is used to drive the fourth NMOS transistor to generate current using voltage. The gate and source of the sixth NMOS transistor are shorted, so that the current difference between the fifth PMOS transistor and the sixth NMOS transistor flows into the second capacitor, generating a slow variable.

4. The artificial neuron circuit according to any one of claims 1 to 3, wherein The volatile threshold switching device is based on metal-insulator switching, metal-conductive filament diffusion, and chalcogenide-based threshold switching devices. Threshold switching devices based on metal-insulator transition employ or Mott insulator materials; Threshold switching devices based on diffusion of conductive metal filaments use Ag or Cu active metal materials to form conductive filaments; threshold switching devices based on chalcogenides use tellurium-based, selenium-based, or sulfide-based OTS materials doped with single or multiple elements; where x > 0.

5. The artificial neuron circuit of claim 3, wherein, The first and second capacitors are either fixed or variable capacitors, with capacitance values ​​ranging from 1fF to 1μF.

6. The artificial neuron circuit of claim 1, wherein, The firing behavior of biological neurons includes regular firing, initiation cluster firing, cluster firing, and rapid firing.

7. The artificial neuron circuit according to claim 1 or 6, wherein The state equation of an artificial neuron circuit is: in, V and U These are membrane potential and slow variable, respectively; For input current; and These are the currents of the third PMOS transistor and the fifth PMOS transistor, respectively; and The currents of the fourth and sixth NMOS transistors; This is the capacitance value of the first capacitor; This is the capacitance value of the second capacitor.

8. An artificial neuron firing method based on the artificial neuron circuit according to claim 3, characterized by, Includes the following steps: Step S1: Under the application of external reset voltage and input current, the membrane potential rises rapidly through the positive feedback loop of the membrane potential generation circuit. When the difference between the membrane potential and the external reset voltage is greater than the threshold voltage of the volatile threshold switching device, the volatile threshold switching device changes from a high-resistance state to a low-resistance state, providing a low-resistance discharge path for the first capacitor, causing the membrane potential to drop rapidly. The current of the first NMOS transistor increases rapidly and then decreases. Through the current mirror circuit, the current corresponding to the fifth PMOS transistor also increases rapidly and then decreases, charging the second capacitor and increasing the slow variable. Step S2: When the difference between the membrane voltage and the external reset voltage is less than the holding voltage of the volatile threshold switching device, the volatile threshold switching device changes from a low-resistance state to a high-resistance state; the membrane potential is continuously charged and reset, the slow variable continues to increase, and the neuron enters the initial stage of firing; Step S3: When the slow variable increases to its maximum value, the current in the fourth NMOS transistor and the sixth NMOS transistor increases, making the rate of change of the membrane potential and the rate of change of the slow variable both less than 0, and the membrane potential and the slow variable decrease. Step S4: During the decrease of the slow variable, the rate of change of the membrane potential is greater than 0, the rate of change of the slow variable is less than 0, the membrane potential rises, the slow variable continues to decrease, and the neuronal firing behavior enters the refractory period. Step S5: The slow variable continues to decrease, and the rate of change of both the membrane potential and the slow variable is greater than 0, causing the membrane potential to be continuously charged and reset. Proceed to step S1 until the artificial neuron stops firing.

9. The artificial neuron firing method of claim 8, wherein, The firing behavior of biological neurons includes regular firing, initiation cluster firing, cluster firing, and rapid firing. The external reset voltage required for rapid firing is greater than that required for cluster firing; the external reset voltage required for cluster firing is greater than that required for initiation cluster firing; and the external reset voltage required for initiation cluster firing is greater than that required for regular firing.

Images