Event trigger type neuron circuit based on volatile memristor

Abstract

The invention relates to an event trigger type neuron circuit based on a volatile memristor. The event trigger type neuron circuit comprises a signal transmission module and a coding module, the signal transmission module is used for converting the received illumination signal into an output voltage and transmitting the output voltage to the coding module; wherein the output voltage is positively correlated with the intensity of the illumination signal; and the coding module is used for obtaining an oscillation coding signal according to the variable quantity of the output voltage. According to the technical scheme, the illumination signal is converted into the voltage, the oscillation coding signal consistent with LIF neuron characteristics is obtained through the variable quantity of the output voltage, and the output voltage can be changed according to the increase and decrease of the identification illumination intensity and is also consistent with the characteristics of an event driving structure; voltage coding output is carried out on illumination changes by utilizing a simple circuit structure, and development of event-driven structure research in brain-like calculation is promoted to a certain extent.

Description

Technical Field

[0001] The present invention relates to the field of electronic information technology, and in particular to an event-triggered neuron circuit based on a volatile memristor. Background Art

[0002] Brain-like computing is a computing method that simulates the electrophysiological process of cooperation between neurons in the human brain. It can be applied to artificial intelligence, pattern recognition, control and other fields. Different from the traditional von Neumann architecture, brain-like computing does not require a large amount of data to be transmitted to each other, and processes more complex information in a more flexible way. Compared with traditional computers, cells in neural networks process a large amount of information at the same time, which makes brain-like computing have great advantages in simulating the human brain to process information.

[0003] Event cameras, also known as event-based cameras or event-driven cameras, are a new type of visual sensing technology. Significantly different from traditional cameras in the way data is collected and represented, event cameras work not by capturing images of the entire scene at a fixed frame rate, but by recording events based on brightness changes at the pixel level. Each pixel in an event camera works independently, and when a change in pixel brightness is detected, the pixel triggers an event. Traditional cameras capture images of the entire scene at fixed time intervals (such as 30 or 60 frames per second). Each frame contains static image information of the entire scene. Every frame of the entire scene is recorded, regardless of whether the scene changes. Each frame contains a lot of redundant information, especially when most of the scene remains unchanged. Limited by the fixed frame rate and relatively low dynamic range, traditional cameras may blur or lose information for fast-moving objects, and are prone to losing details under strong lighting.

[0004] The individual pixels of an event camera are used to detect brightness changes and generate events when these changes occur. This structure enables the event camera to respond to brightness changes with extremely high temporal resolution while only generating data related to the changes, which greatly reduces the amount of data and processing requirements. The traditional event camera pixel consists of a mixed digital-analog circuit, including a photosensitive element, a comparator, a threshold adjuster, an addressing logic, a timestamp generator, and an event output structure. Event cameras have broad application potential in areas such as autonomous driving, robot vision, aerospace, and high-speed video capture, especially in scenarios that require high temporal resolution, low latency, and high dynamic range visual information processing. Combining event-driven with brain-like computing has a wide range of applications in scenarios such as dynamic image recognition, speed detection, and motion prediction, and the overall circuit power consumption can be further reduced. However, in current research, the structure of a single pixel of an event camera is complex, the layout area of ​​the design and manufacturing is large, and it is inconvenient to test the performance of a single pixel. The differences in different structures directly or indirectly affect the overall performance, which to a certain extent limits the further reduction of power consumption and further simplification of the structure, hindering the application of event cameras in brain-like computing. Summary of the invention

[0005] In order to solve the above problems existing in the prior art, the present invention provides an event-triggered neuron circuit based on a volatile memristor.

[0006] According to a first aspect of an embodiment of the present invention, there is provided an event-triggered neuron circuit based on a volatile memristor, the event-triggered neuron circuit based on a volatile memristor comprising: a signal transmission module and an encoding module;

[0007] The signal transmission module is used to convert the received light signal into an output voltage, and transmit the output voltage to the encoding module; wherein the output voltage is positively correlated with the intensity of the light signal;

[0008] The encoding module is used to obtain an oscillation encoding signal according to the change amount of the output voltage.

[0009] Optionally, the signal transmission module includes: an external gate power supply, an external drain power supply, a photoelectric thin-film transistor and a voltage transmission resistor; the positive electrode of the external gate power supply is connected to the gate of the photoelectric thin-film transistor, the positive electrode of the external drain power supply is connected to the drain of the photoelectric thin-film transistor, the source of the photoelectric thin-film transistor is connected to the first end of the voltage transmission resistor, the negative electrode of the external gate power supply is grounded, the negative electrode of the external drain power supply is grounded, and the second end of the voltage transmission resistor is grounded.

[0010] Optionally, the signal transmission module is further used to obtain a current value flowing through a voltage-transmitting resistor according to the light intensity currently received by the photoelectric thin-film transistor; and is further used to obtain the output voltage according to the current value.

[0011] Optionally, the gate of the photoelectric thin film transistor is used to detect light intensity.

[0012] Optionally, the encoding module includes: a differential capacitor and a volatile memristor; a first end of the differential capacitor is connected to the output voltage, a second end of the differential capacitor is connected to the first end of the volatile memristor, and a second end of the volatile memristor is grounded.

[0013] Optionally, the encoding module is also used to obtain the charging and discharging current of the differential capacitor according to the output voltage; it is also used to obtain the first voltage of the volatile memristor in a high-resistance state and the second voltage of the volatile memristor in a low-resistance state according to the charging and discharging current; it is also used to obtain the oscillation coding signal according to the first voltage and the second voltage.

[0014] The technical solution provided by the present invention may include the following beneficial effects:

[0015] Through the above technical scheme, the signal transmission module is used to convert the light signal into voltage, and the change in output voltage is used to obtain an oscillation coding signal consistent with the characteristics of LIF neurons. The output voltage can change according to the increase and decrease in the identification of light intensity, which is also consistent with the characteristics of the event-driven structure. It realizes the voltage coding output of light changes using a simple circuit structure, which to a certain extent promotes the development of event-driven structure research in brain-like computing.

[0016] Other features and advantages of the present invention will be described in detail in the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the following specific embodiments, they are used to explain the present invention, but do not constitute a limitation of the present invention. In the accompanying drawings:

[0018] Figure 1 is a schematic diagram of an event-triggered neuron circuit based on a volatile memristor according to an exemplary embodiment.

[0019] Figure 2 It is a structural diagram of a signal transmission module according to an exemplary embodiment.

[0020] Figure 3 It is a schematic diagram of the structure of a coding module according to an exemplary embodiment.

[0021] Figure 4 It is a schematic diagram of the structure of an event-triggered neuron circuit based on a volatile memristor according to an exemplary embodiment.

[0022] Figure 5 The diagram is an input and output diagram of an event-triggered neuron circuit based on a volatile memristor according to an exemplary embodiment.

[0023] Figure 6 The diagram is a schematic diagram of an output of an event-triggered neuron circuit based on a volatile memristor according to an exemplary embodiment.

[0024] Figure 7 The diagram is a schematic diagram of an output of an event-triggered neuron circuit based on a volatile memristor according to an exemplary embodiment.

[0025] Figure 8 is an output schematic diagram of another event-triggered neuron circuit based on a volatile memristor according to an exemplary embodiment. DETAILED DESCRIPTION

[0026] Figure 1 is a schematic diagram of an event-triggered neuron circuit based on a volatile memristor according to an exemplary embodiment. Figure 1 As shown, the event-triggered neuron circuit 10 based on volatile memristor includes: a signal transmission module 101 and an encoding module 102;

[0027] The signal transmission module 101 is used to convert the received light signal into an output voltage Vout1, and transmit the output voltage Vout1 to the encoding module 102; wherein the output voltage Vout1 is positively correlated with the intensity of the light signal;

[0028] The encoding module 102 is used to obtain an oscillation encoding signal Vout2 according to the variation of the output voltage Vout1.

[0029] Optionally, Figure 2 is a schematic diagram of a structure of a signal transmission module according to an exemplary embodiment. Figure 2 As shown, the signal transmission module 101 includes: an external gate power supply Vg, an external drain power supply Vdd, a photoelectric thin film transistor TFT and a voltage transmission resistor R; the positive electrode of the external gate power supply Vg is connected to the gate of the photoelectric thin film transistor TFT, the positive electrode of the external drain power supply Vdd is connected to the drain of the photoelectric thin film transistor TFT, the source of the photoelectric thin film transistor TFT is connected to the first end of the voltage transmission resistor R, the negative electrode of the external gate power supply Vg is grounded, the negative electrode of the external drain power supply Vdd is grounded, and the second end of the voltage transmission resistor R is grounded.

[0030] It can be understood that the external gate power supply Vg and the external drain power supply Vdd are adjustable voltage sources. The present invention utilizes the external gate power supply Vg to simulate changes in light intensity. The external gate power supply Vg and the external drain power supply Vdd usually put the photoelectric thin film transistor TFT in a saturated working state; the resistance value of the voltage transmission resistor R is adjustable. When the resistance value of the voltage transmission resistor R is larger, the output voltage Vout1 is larger. When the resistance value of the voltage transmission resistor R is smaller, the output voltage Vout1 is smaller.

[0031] Optionally, the signal transmission module is further used to obtain a current value flowing through the voltage transmission resistor R according to the light intensity currently received by the photoelectric thin film transistor TFT; and is further used to obtain an output voltage Vout1 according to the current value.

[0032] Optionally, the gate of the photoelectric thin film transistor TFT is used to detect light intensity.

[0033] It can be understood that the present invention uses a photoelectric thin film transistor TFT to sense light intensity, forming a signal transmission module, converting the light intensity into an output voltage Vout1 proportional to it, and transmitting it to the encoding module. The photoelectric thin film transistor is an active electronic device. Unlike ordinary transistors, the photoelectric thin film transistor is small in size and can sense light intensity. When illuminated or de-illuminated, photogenerated carriers are generated or recombined in the conductive channel, thereby increasing or decreasing the channel current. A resistor is connected in series at its source level, and the increased current increases or decreases the potential difference on the resistor, causing the output voltage to increase or decrease, thereby having the function of converting light intensity into voltage and transmitting it.

[0034] Optionally, Figure 3 is a schematic diagram of a structure of a coding module according to an exemplary embodiment. Figure 3 As shown, the encoding module 102 includes: a differential capacitor C and a volatile memristor TS; the first end of the differential capacitor C is connected to the output voltage Vout1, the second end of the differential capacitor C is connected to the first end of the volatile memristor TS, and the second end of the volatile memristor TS is grounded.

[0035] Optionally, the encoding module is also used to obtain the charging and discharging current of the differential capacitor C according to the output voltage Vout1; it is also used to obtain the first voltage of the volatile memristor TS in a high-resistance state and the second voltage of the volatile memristor TS in a low-resistance state according to the charging and discharging current; it is also used to obtain the oscillation encoding signal Vout2 according to the first voltage and the second voltage.

[0036] It can be understood that the differential capacitor C is a differentiator. The capacitor is a passive electronic component, and the amount of charge stored inside it is proportional to the potential difference between its positive and negative poles. When the potential difference between the two ends increases, the differential capacitor C will be in a charging state, and when the potential difference between the two ends decreases, the differential capacitor C will be in a discharging state. During the charging and discharging process, the differential capacitor C can be regarded as a short-circuit state, and can be regarded as an open circuit state after charging is completed. Therefore, the differential capacitor C has the function of detecting signal changes; the volatile memristor TS is used to simulate the imitation electric pulse behavior of the LIF neuron. The volatile memristor TS is a passive electronic component. It is similar to a resistor and can generate and maintain current through a device. However, unlike a resistor, the volatile memristor TS can still "remember" the amount of charge that passed through it before after the power supply is disconnected. There are two important parameters in the volatile memristor TS, namely the threshold voltage Vth and the holding voltage Vhold. The threshold voltage refers to the voltage value required for the memristor to switch from one resistance state to another resistance state. When the applied voltage exceeds this threshold, the resistance state of the volatile memristor TS changes; for example, when the applied voltage exceeds the threshold voltage Vth, the volatile memristor TS switches from a high resistance state to a low resistance state, and the resistance value of the volatile memristor TS changes to a lower resistance value. The holding voltage Vhold refers to the minimum voltage required for the volatile memristor TS to maintain the low resistance state after it switches to the low resistance state. Once the volatile memristor TS switches to a new resistance state through a signal exceeding the threshold voltage Vth, it can maintain its resistance state above the threshold voltage as long as no voltage is applied that is sufficient to trigger the state change. Due to special reasons such as the manufacturing process, the volatile memristor has parasitic capacitance inside, which is called membrane capacitance by analogy with nerve cells. When a certain voltage is applied to it, it can produce oscillation behavior similar to the output of neuron signals.

[0037] It is worth mentioning that due to the participation of the differential capacitor C, this process only occurs when the light intensity changes. If the light intensity remains unchanged, or the change in voltage caused by the change in light intensity is not sufficient to cause the volatile memristor to switch its resistance state through the detection of the differential capacitor C, then no oscillation encoding signal Vout2 will be generated.

[0038] It can be understood that the capacitance of the differential capacitor C is set according to actual conditions. The first end of the differential capacitor C is used to receive the output voltage Vout1 of the signal transmission module 101. The current flowing through the volatile memristor TS is obtained according to the change of the output voltage Vout1. The larger the capacitance of the differential capacitor C, the slower the charging speed of the output voltage Vout1 to the differential capacitor C, and the longer the duration of the current flowing through the volatile memristor TS. The smaller the capacitance of the differential capacitor C, the faster the charging speed of the output voltage Vout1 to the differential capacitor C, and the shorter the duration of the current flowing through the volatile memristor TS. The volatile memristor TS can be regarded as a resistor that switches back and forth between different resistance states. The current flowing through the volatile memristor TS multiplied by the resistance of the volatile memristor TS in different states can obtain the voltage across the volatile memristor TS. By comparing the voltage across the volatile memristor TS with the threshold voltage Vth and the holding voltage Vhold of the volatile memristor TS, the resistance state of the volatile memristor TS at the next moment can be obtained. The output Vout2 of the encoding module can be obtained by the spontaneous switching of the voltage across the volatile memristor TS.

[0039] In one embodiment, in the initial state of the event-triggered neuron circuit based on the volatile memristor TS, the volatile memristor TS is in a high-resistance state, with a resistance value equal to its high-resistance resistance, an external gate voltage source Vg and an external drain voltage source Vdd are given a certain value, and the photoelectric thin-film transistor TFT is in a saturation region or a linear region. At this time, there is a certain voltage across the voltage-transmitting resistor R, but the voltage remains unchanged, and there is a certain voltage across the differential capacitor C, but at this time there is no current passing through the differential capacitor C, so the voltage across the volatile memristor TFT is 0; due to the photoelectric The light intensity received by the gate of the thin film transistor TFT changes. If the light intensity increases, photogenerated carriers will be generated inside the photoelectric thin film transistor TFT, thereby increasing the current passing through the photoelectric thin film transistor TFT when the external gate voltage source Vg and the external drain voltage source Vdd remain unchanged. The resistance value of the voltage transmission resistor R remains unchanged, so the voltage across the voltage transmission resistor R will increase with the increase of light, and the potential of one end of the differential capacitor C connected to the signal transmission module 101 will increase accordingly, and the differential capacitor C will be in the charging process until the differential capacitor C is connected to the signal transmission module 101. The potential of one end of the module 101 is equal to the potential of one end of the voltage-transmitting resistor R connected to the source of the photoelectric thin-film transistor TFT. During the charging process, a current will flow through the volatile memristor TS through the differential capacitor C. At this time, the voltage across the two ends of the volatile memristor TS is determined by the high resistance state and the current flowing through the differential capacitor C. If the voltage across the two ends of the volatile memristor TS is greater than the threshold voltage during the charging process, the volatile memristor TS will immediately change to a low resistance state. At this time, the voltage across the two ends of the volatile memristor TS is determined by the low resistance state and the current flowing through the differential capacitor C. If the voltage across the two ends of the volatile memristor TS is greater than the threshold voltage during the charging process, the volatile memristor TS will immediately change to a low resistance state. At this time, the voltage across the two ends of the volatile memristor TS is determined by the low resistance state and the current flowing through the differential capacitor C. If the voltage across the volatile memristor TS is less than the holding voltage, the volatile memristor TS will immediately change to a high-resistance state. Therefore, during the charging process of the differential capacitor C, the volatile memristor TS will continuously change the state of resistance, so that the end of the volatile memristor TS connected to the differential capacitor C has an oscillating output similar to neuron coding; it is worth mentioning that during the charging process of the differential capacitor C, the direction of the current is the positive direction, and the oscillation coding signal Vout2 is the positive output; it is understandable that when the charging process of the differential capacitor C ends and the current becomes 0, there is no oscillating output.If the light intensity decreases, the photogenerated carriers will be recombined inside the photoelectric thin film transistor TFT, thereby reducing the current passing through the photoelectric thin film transistor TFT when the external gate voltage source Vg and the external drain voltage source Vdd remain unchanged, and the resistance value of the pressure transfer resistor R remains unchanged. Therefore, the voltage across the pressure transfer resistor R will decrease as the light intensity decreases, and the potential of one end of the differential capacitor C connected to the signal transmission module 101 will decrease accordingly. The differential capacitor C will be in a discharge process until the potential of one end of the differential capacitor C connected to the signal transmission module 101 is equal to the potential of one end of the pressure transfer resistor R connected to the source of the photoelectric thin film transistor TFT. During the discharge process, a current will flow through the differential capacitor C and the volatile memristor T S, at this time, the voltage across the volatile memristor TS is determined by the high resistance state and the current flowing through the differential capacitor C. If the voltage across the volatile memristor TS is greater than the threshold voltage during the discharge process, the volatile memristor TS will immediately change to a low resistance state. At this time, the voltage across the volatile memristor TS is determined by the low resistance state and the current flowing through the differential capacitor C. If the voltage across the volatile memristor TS is less than the holding voltage during the discharge process, the volatile memristor TS will immediately change to a high resistance state. Therefore, during the discharge process of the differential capacitor C, the volatile memristor TS will continuously change the state of resistance, so that the end of the volatile memristor TS connected to the differential capacitor C has an oscillation output similar to neuron coding. During the discharge process of the differential capacitor C, the direction of the current is negative, and the oscillation coding signal Vout2 is a reverse output. It can be understood that when the charging process of the differential capacitor C ends and the current becomes 0, there is no oscillation output.

[0040] Among them, the pressure-transmitting resistor R plays two roles in the charging and discharging process of the differential capacitor C: during the charging process, the size of the pressure-transmitting resistor R can directly determine the potential of one end of the differential capacitor C connected to the signal transmission module 101; during the discharging process, the size of the pressure-transmitting resistor R can directly determine the discharge speed of the differential capacitor C; therefore, in the implementation process, it becomes particularly important to select a suitable resistance value of the pressure-transmitting resistor R.

[0041] In one embodiment, Figure 4 is a schematic diagram of a structure of an event-triggered neuron circuit based on a volatile memristor according to an exemplary embodiment. Figure 4 As shown, according to Figure 2 and Figure 3 , the structure of the signal transmission module and the structure of the encoding module can be connected to an event-triggered neuron circuit based on a volatile memristor.

[0042] In one embodiment, in order to better demonstrate the effect of the present invention, a simulation experiment can be performed based on the present invention. Among them, the low resistance state of the volatile memristor TS is 2kΩ, the high resistance state is 80kΩ, the pressure transfer resistor R is 22k, the differential capacitor C is 33n, the external drain power supply Vdd is 10v, the gate pulse of the photoelectric thin film transistor TFT is 10v, the pulse period is 200ms, the high level is 10ms, the rise time is 1ms, and the fall time is 1ms. The high level of the external gate power supply Vg is 10v, the low level is 0V, the pulse period is 200ms, the high level is 10ms, the rise time is 1ms, and the fall time is 1ms. Figure 5 is an input and output schematic diagram of an event-triggered neuron circuit based on a volatile memristor according to an exemplary embodiment. Figure 6 is a schematic diagram showing an output of an event-triggered neuron circuit based on a volatile memristor according to an exemplary embodiment. Figure 5 and Figure 6 As shown, green is the output of the oscillation coding signal Vout2, and blue is the change of the external gate power supply Vg. Figure 7 is an output schematic diagram of an event-triggered neuron circuit based on a volatile memristor according to an exemplary embodiment. Figure 8 is an output schematic diagram of another event-triggered neuron circuit based on a volatile memristor according to an exemplary embodiment. Figure 7 It is the positive output corresponding to the rising edge of the gate pulse of the photoelectric thin film transistor TFT. Figure 8 It is the negative output corresponding to the falling edge of the gate pulse of the photoelectric thin film transistor TFT.

[0043] Through the above technical solution, the light intensity is detected by the photoelectric thin film transistor TFT to obtain an output voltage Vout1 proportional to the light intensity. The change of light affects the change of the output voltage Vout1. The change of the output voltage Vout1 is detected by the differential capacitor C to obtain the current flowing through the volatile memristor TS. When the volatile memristor TS is in a high resistance state, the current passing through the volatile memristor TS will cause the volatile memristor TS to be in a state of high voltage at both ends. When this voltage exceeds the threshold voltage of the volatile memristor TS, the volatile memristor TS will be converted to a low resistance state, and at this time, the volatile memristor TS will be converted to a low resistance state. The current of the resistor TS increases accordingly, but at this time the voltage across the volatile memristor TS is not enough to maintain the low-resistance state of the volatile memristor TS, thereby causing the volatile memristor TS to switch back and forth between the high and low resistance states, and obtaining an oscillation coding signal Vout2. The output oscillation coding signal Vout2 is consistent with the characteristics of the LIF neuron circuit, and is only output when a change in light intensity is detected. It has an event-triggered characteristic, and realizes the use of a simple circuit system to complete an event-driven structure for brain-like computing, which to a certain extent promotes the development of research on brain-like computing structures with event-driven characteristics.

[0044] The preferred embodiments of the present invention are described in detail above in conjunction with the accompanying drawings. However, the present invention is not limited to the specific details in the above embodiments. Within the technical concept of the present invention, a variety of simple modifications can be made to the technical solution of the present invention, and these simple modifications all belong to the protection scope of the present invention.

[0045] It should also be noted that the various specific technical features described in the above specific embodiments can be combined in any suitable manner without contradiction. In order to avoid unnecessary repetition, the present invention will not further describe various possible combinations.

[0046] In addition, various embodiments of the present invention may be arbitrarily combined, and as long as they do not violate the concept of the present invention, they should also be regarded as the contents disclosed by the present invention.

Claims

1. An event-triggered neuron circuit based on a volatile memristor, characterized in that: The event-triggered neuron circuit based on volatile memristor includes: a signal transmission module and a coding module; The signal transmission module is used to convert the received light signal into an output voltage, and transmit the output voltage to the encoding module; wherein the output voltage is positively correlated with the intensity of the light signal; The encoding module is used to obtain an oscillation encoding signal according to the change amount of the output voltage.

2. The event-triggered neuron circuit based on volatile memristor according to claim 1, characterized in that: The signal transmission module includes: an external gate power supply, an external drain power supply, a photoelectric thin-film transistor and a voltage transmission resistor; the positive electrode of the external gate power supply is connected to the gate of the photoelectric thin-film transistor, the positive electrode of the external drain power supply is connected to the drain of the photoelectric thin-film transistor, the source of the photoelectric thin-film transistor is connected to the first end of the voltage transmission resistor, the negative electrode of the external gate power supply is grounded, the negative electrode of the external drain power supply is grounded, and the second end of the voltage transmission resistor is grounded.

3. The event-triggered neuron circuit based on volatile memristor according to claim 2, characterized in that: The signal transmission module is further used to obtain the current value flowing through the voltage transmission resistor according to the light intensity currently received by the photoelectric thin film transistor; and is also used to obtain the output voltage according to the current value.

4. The event-triggered neuron circuit based on volatile memristor according to claim 3, characterized in that: The gate of the photoelectric thin film transistor is used to detect light intensity.

5. The event-triggered neuron circuit based on volatile memristor according to claim 1, characterized in that: The encoding module includes: a differential capacitor and a volatile memristor; the first end of the differential capacitor is connected to the output voltage, the second end of the differential capacitor is connected to the first end of the volatile memristor, and the second end of the volatile memristor is grounded.

6. The event-triggered neuron circuit based on volatile memristor according to claim 5, characterized in that: The encoding module is also used to obtain the charging and discharging current of the differential capacitor according to the output voltage; it is also used to obtain the first voltage of the volatile memristor in a high-resistance state and the second voltage of the volatile memristor in a low-resistance state according to the charging and discharging current; it is also used to obtain the oscillation encoding signal according to the first voltage and the second voltage.

Images