Emotion memory generation system based on memristor

Abstract

The invention discloses an emotion memory generation system based on a memristor, and belongs to the field of integrated circuit design. The system comprises a brain emotion learning model and a hippocampus model. The brain emotion learning model comprises a thalamus module, a sensory cortex module, an orbital frontal cortex module and an amygdala kernel module; the hippocampus model comprises a DG module, a CA3 module and a CA1 module; under the stimulation of the emotional signals, the thalamus module firstly preprocesses the emotional signals, then the sensory cortex module and the orbital frontal lobe module integrate and process the preprocessed emotional signals, and the processed signals are respectively transmitted to the CA3 module and the amygdala kernel module to generate output signals representing emotions; under the stimulation of a learning signal, the DG module firstly preprocesses the learning signal, then the CA3 region is responsible for receiving the learning signal and generating an output signal representing short-term memory, and the output signal is transmitted to the CA1 region; and after receiving the short-term memory signal, the CA1 area converts short-term memory into long-term memory and stores the long-term memory.

Description

Technical Field

[0001] The present invention belongs to the field of integrated circuit design, and particularly relates to an emotion memory generation system based on memristors. Background Art

[0002] As artificial intelligence (AI) gradually approaches the level of the human brain in terms of speed and accuracy, we have come to recognize the complexity of the human brain structure and its advanced cognitive functions, such as memory, learning, decision-making, and emotional expression. One of the main trends driving the continuous progress of AI is the research and simulation of the efficiency and capabilities of the brain in information processing.

[0003] Multiple studies have proposed that implementing some or all of the computational functions of the brain through hardware circuits will greatly improve the computational speed and may open a new chapter in the field of artificial intelligence. However, the current mainstream hardware design still follows the von Neumann architecture, which has inherent limitations in information processing and storage, posing unprecedented challenges to the design and implementation of emotion memory generation networks. Against this background, hardware circuits based on memristors have emerged as an innovative solution to break the shackles of the von Neumann architecture. Memristor technology has achieved a deep integration of storage and computational functions, significantly reducing the need for data transmission and thus greatly improving computational efficiency.

[0004] The development of human-robot interaction robots faces problems such as difficulties in emotion recognition and expression, weak adaptability and learning ability, and complex interaction design. It is hoped that robots can perform emotional learning like humans. The learning and generation of emotions are not only a key way of information exchange but also the basis of human psychological and physiological activities. Future intelligent robots are expected to be constructed using the concept of combining cognition and emotion, and such robots will be able to show empathy in human-robot interaction. Summary of the Invention

[0005] Aiming at the deficiencies of the prior art, the purpose of the present invention is to provide an emotion memory generation system based on memristors, which solves the problems in the prior art.

[0006] The purpose of the present invention can be achieved by the following technical solutions:

[0007] An emotion memory generation system based on memristors includes a brain emotion learning model and a hippocampus model; the brain emotion learning model includes: a thalamus module, a sensory cortex module, an orbitofrontal cortex module, and an amygdala module; the hippocampus model includes: a DG module, a CA3 module, and a CA1 module;

[0008] Under the stimulation of emotional signals, the thalamus module first preprocesses the emotional signals, and then the sensory cortex module and the orbitofrontal module integrate and process the preprocessed emotional signals, and transmit the processed signals to the CA3 module and the amygdala module respectively to generate output signals representing emotions;

[0009] Under the stimulation of learning signals, the DG module first preprocesses the learning signals, and then the CA3 area is responsible for receiving the learning signals and generating output signals representing short-term memory, which are transmitted to the CA1 area; after receiving the short-term memory signals, the CA1 area converts the short-term memory into long-term memory and stores it.

[0010] Furthermore, the preprocessing of emotional signals by the thalamus module includes: receiving external emotional signals, dividing the emotional signals into positive and negative polarity signals through the thalamus module, where the positive polarity signals represent positive emotional signals and the negative polarity signals represent negative emotional signals, and selecting emotional signals or forgetting signals to be transmitted to the sensory cortex module and the orbitofrontal module through control signals.

[0011] Furthermore, the integration and processing of emotional signals by the sensory cortex module and the orbitofrontal module include: the thalamus module can receive multiple emotional signals and preprocess the multiple emotional signals into multiple positive polarity signals and multiple negative polarity signals. The sensory cortex module first integrates the signals from the thalamus module and transmits the integrated signals to the CA3 module and the orbitofrontal module respectively. Subsequently, the orbitofrontal module extracts specific features from the sensory cortex module and evaluates their emotional significance.

[0012] Furthermore, the thalamus module includes two circuits. One circuit includes: two rectifier operational amplifiers, two transmission gate switches S1, S2 and an inverter. The first rectifier operational amplifier includes: operational amplifier A1-1, resistors R1-R3 and diode D1. The second rectifier operational amplifier includes: operational amplifier A1-2, resistors R4-R5 and diode D2. The transmission gates S1, S2 are controlled by the control signal VC to select the Vin signal or Vre to be transmitted to the subsequent module. Among them, the Vin signal will be processed into positive and negative polarity signals by operational amplifiers A1-1, A1-2, resistors R1-R5 and two diodes D1, D2;

[0013] Another circuit in the thalamus module includes operational amplifiers A1-3, A1-4 and resistors R6-R9 and transmission gate switches S3, S4, and can also process two emotional signals Vin simultaneously.

[0014] Furthermore, the sensory cortex module and the orbitofrontal cortex module include: a circuit for processing signals VG1, VG2, VForget1 and a circuit for processing signals VB1, VB2, VForget2;

[0015] The circuit for processing signals VG1, VG2, and VForget1 includes: an adder, a differential amplifier, and a negative feedback integration loop; the adder includes operational amplifier A2-1 and resistors R10-R13, the differential amplifier includes non-volatile memristor NVM1, operational amplifier A2-2, and resistor R14, and the negative feedback integration loop includes operational amplifiers A2-3, A2-4, A2-5, resistors R15-R19, and capacitor C1; the input signal from the thalamus module is integrated by the adder composed of operational amplifier A2-1 and resistors R10-R13, and the integrated emotion signal is input into the CA3 module. At the same time, it is also input into the orbitofrontal module through the differential amplifier; the orbitofrontal module receives signals from the sensory cortex module and is transformed into a stable continuous signal VOG through the negative feedback integration loop composed of operational amplifiers A2-3, A2-4, A2-5, resistors R15-R19, and capacitor C1, and is transmitted to the amygdala module;

[0016] The circuit for processing signals VB1, VB2, and VForget2 includes operational amplifiers A2-6, A2-7, A2-8, A2-9, A2-10, resistors R20-R29, and capacitor C2; it is used to process negative-polarity emotion signals and generate a continuous signal VOB for transmission to the amygdala module;

[0017] The output of the adder in the sensory cortex module is sent to the CA3 module, and the output of the orbitofrontal cortex module is sent to the amygdala module.

[0018] Further, the amygdala module includes: three inverting adders, two comparators B3-1 and B3-2, two followers A3-3, A3-5, and two PMOS switch transistors; the first adder includes operational amplifier A3-1, resistors R30, R31, and volatile memristor VM1, the second adder includes operational amplifier A3-4, resistors R32, R33, and volatile memristor VM2; the third adder includes operational amplifier A3-2, non-volatile memristors NVM3, NVM4, and resistor R34; the first adder and the second adder receive signals from the orbitofrontal cortex module and the CA1 module, and the third adder receives the outputs of the first adder and the second adder; the two comparators B3-1 and B3-2 are respectively connected to the followers A3-3 and A3-5; the outputs of the two followers are respectively connected to the same PMOS switch transistor; the source of the PMOS switch transistor is connected to the follower, and the drain is connected to the CA1 module; the output of the amygdala module is controlled by VETM and input into the CA1 module;

[0019] The amygdala module receives the emotional signals processed by the orbitofrontal cortex module. VOG and VOB simultaneously receive the long-term memories VLM+ and VLM- generated from the CA1 region. The positive polarity signal VOG and VLM+ are received by the first adder, and the negative polarity signal VOB and VLM- are received by the second adder. The outputs of the first adder and the second adder are integrated by the third adder, and the integrated signal of the third adder generates the emotional arousal signals EG1 and EB1 through the comparison module. The generated emotional arousal signals are transmitted to the CA1 module through two PMOS switches controlled by the VETM signal respectively.

[0020] Further, the DG module includes: a rectifying operational amplifier, a transmission gate switch S5, and an inverter; wherein the rectifying operational amplifier includes: operational amplifiers A1-5, A1-6, resistors R35-39, and diodes D5 and D6; the control signal controls the transmission gate S5. When the transmission gate is turned on, the input signal Vin is preprocessed into a positive polarity learning signal VL+ and a negative polarity learning signal VL- through the operational amplifiers A1-5, A1-6, resistors R35-39, and D5, D6.

[0021] The DG module divides the Vin signal into two signals VL+ and VL- of positive and negative polarities and inputs them into the CA3 module.

[0022] Further, the CA3 module includes: an emotion influence circuit, a learning signal circuit for processing positive signals, and a learning signal circuit for processing negative signals.

[0023] The emotion influence circuit includes: a reverse adder, a transmission gate switch S6, two comparators, and two followers; wherein the reverse adder includes: operational amplifier A4-1 and resistors R40-R42; the comparator B4-2 and the follower A4-5, and the comparator B4-3 and the follower A4-6 constitute a comparison module.

[0024] The output of the adder in the sensory cortex module is sent into the CA3 module through the emotion influence circuit; when stimulated by the learning signal VL+ or VL-, the VMCMG signal and the VMCMB signal from the sensory cortex module are integrated by the reverse adder and then input into the learning signal circuit; at the same time, the reverse adder is also connected to the comparison module, and by comparing the output of the reverse adder, the emotion valence signals EG2 and EB2 are obtained; the emotion valence signals EG2 and EB2 and the emotional arousal signals EG1 and EB1 constitute the emotion space representing emotions.

[0025] The learning signal circuit for processing the positive signal includes: an inverting adder, an inverting proportional amplifier, and a comparison module composed of a comparator and a follower. The inverting adder includes: operational amplifier A4-2, volatile memristor VM3, and resistors R43-R44. The inverting proportional amplifier includes: operational amplifier A4-3, resistors R45 and R46. The comparison module includes: comparator B4-1 and follower A4-4.

[0026] The positive-polarity learning signal VL+ processed by the DG module, through the inverting adder, first performs an addition operation with the signal from the emotional influence learning circuit. Then, through the comparison module, if the generated signal VSM+ is produced, it represents that the short-term memory has been generated at this time. Subsequently, the VSM+ signal is transmitted to the CA1 module, and long-term memory is generated in the CA1 module.

[0027] The learning signal circuit for processing the negative signal includes: volatile memristor VM4, operational amplifiers A4-7, A4-8, A4-9, and resistors R47-R50, which are used to process the negative-polarity learning signal VL-, generate VSM-, and input it into the subsequent CA1 module to generate long-term memory.

[0028] Furthermore, the CA1 module includes: a circuit for processing the positive signal VSM+, a circuit for processing the negative signal VSM-, and a control circuit composed of four transistors.

[0029] The control circuit: includes two pairs of NMOS and POMS transistors; when the control signal Vctr is at a high level, the control circuit will transmit a recall signal Vread through R52 and R56. When Vctr is at a low level, the forgetting signal Vf will be transmitted through R52 and R56.

[0030] The circuit for processing the positive signal VSM+ includes: an inverting adder, an inverting amplifier, and a comparison module composed of a comparator and a follower. The inverting adder includes: operational amplifier A5-1, volatile memristors VM5, VM6, and resistors R51-R53. The inverting amplifier includes: operational amplifier A5-2, non-volatile memristor NVM5, and resistor R54. The comparison module includes comparator A5-3, follower A5-4, and a PMOS switch.

[0031] The inverting adder receives the emotional arousal signals EG1 and EB1 from the amygdala through VM5 and VM6, receives the signal VSM+ from the CA3 module through resistor R51, and receives the signal from the control circuit through R52. The output of the inverting adder is input into the comparison module through the inverting proportional amplifier. The generated signal LM+ representing long-term memory is transmitted to the amygdala module through the PMOS switch controlled by the control signal VMTE.

[0032] The circuit for processing the negative signal VSM- includes: an inverting adder, an inverting amplifier, and a comparison module composed of a comparator and a follower; where the inverting adder includes: operational amplifier A5-5, volatile memristors VM7, VM8, and resistors R55 - R57; the inverting amplifier includes: operational amplifier A5-6, non-volatile memristor NVM5, and resistor R58; the comparison module includes comparator A5-7, follower A5-8, and a PMOS switch; it is used to receive the VSM- signal from the CA3 module, the EG1 and EB1 signals from the amygdala, and the signal from the control circuit, and generate the signal LM- representing long-term memory, which is transmitted to the amygdala module through the PMOS switch controlled by VMTE;

[0033] The output of CA1 is controlled by the VMTE signal and input into the amygdala module.

[0034] A robot includes the above-mentioned memristor-based emotional memory generation system.

[0035] Advantages of the present invention:

[0036] 1. The invention proposes a bionic circuit for humans to process emotions and memories. Based on the biological mechanism of the hippocampus - amygdala mutual encoding, it characterizes the complex relationship of the interaction between the generation of emotions and memories in the human brain under multi-input signals, providing innovative insights into the emotion generation circuit and its application potential in emotional robots.

[0037] 2. A circuit for generating emotional memory based on memristor technology is proposed. This circuit is significantly different from the traditional von Neumann architecture. It ingeniously integrates the storage and computing functions into one, thus significantly improving the speed and efficiency of information processing. This design not only optimizes the data flow but also reduces the energy consumption, providing a more efficient hardware foundation for simulating the complexity and dynamics of human emotional memory. Description of the Drawings

[0038] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, for those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0039] Figure 1 It is a schematic diagram of the overall structure of the emotional memory production system of the present invention;

[0040] Figure 2 It is a schematic diagram of the circuit connection structure of each module in the emotional memory production system of the present invention;

[0041] Figure 3It is a schematic diagram of the circuit structure of the thalamus module of the present invention;

[0042] Figure 4 It is a schematic diagram of the circuit structure of the sensory cortex module and the orbitofrontal cortex module of the present invention;

[0043] Figure 5 It is a schematic diagram of the circuit structure of the amygdala module of the present invention;

[0044] Figure 6 It is a schematic diagram of the circuit structure of the DG module of the present invention;

[0045] Figure 7 It is a schematic diagram of the circuit structure of the CA3 module of the present invention;

[0046] Figure 8 It is a schematic diagram of the circuit structure of the CA1 module of the present invention;

[0047] Figure 9 It is a simulation result diagram of the emotion generation of the present invention;

[0048] Figure 10 It is a simulation result diagram of the influence of emotion on the learning rate of the present invention;

[0049] Figure 11 It is a simulation result diagram of the emotion-induced memory arousal of the present invention;

[0050] Figure 12 It is a simulation result diagram of the influence of memory learning on emotion experience of the present invention;

[0051] Figure 13 It is a simulation result diagram of the application of the present invention in robot applications. Detailed implementation manners

[0052] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0053] Embodiment 1

[0054] As Figure 1 and Figure 2 shown, a memristor-based emotion memory generation system includes a brain emotion learning model (BEL model) and a hippocampal model; wherein the brain emotion learning model includes: a thalamus module, a sensory cortex module, an orbitofrontal cortex module, and an amygdala module; the hippocampal model includes: a DG module, a CA3 module, and a CA1 module;

[0055] Under the stimulation of emotional signals, the thalamus module first preprocesses the emotional signals to receive external signal stimuli, classifies the external stimulus signals into positive emotional signals and negative emotional signals, and then the sensory cortex module integrates and processes the preprocessed emotional signals, and transmits the integrated signals to the CA3 module of the hippocampus module and the orbitofrontal cortex module respectively. In the CA3 module, an emotional valence signal is generated through the emotional influence learning circuit. At the same time, the orbitofrontal cortex module further processes the signals from the sensory cortex, generates a continuous signal and transmits it to the amygdala module, and finally an output signal representing emotional arousal is generated in the amygdala module; under the stimulation of learning signals, the DG module first preprocesses the learning signals, processes the external signals into positive learning signals and negative learning signals, and then the CA3 area is responsible for receiving the learning signals and generating an output signal representing short-term memory, which is transmitted to the CA1 area; after receiving the short-term memory signal, the CA1 area converts the short-term memory into long-term memory and stores it.

[0056] Under the combined action of emotional signals and learning signals, the system can generate emotions, and the generated emotions and the input emotional signals will promote or inhibit the learning rate. The designed circuit can automatically identify the input signals and generate corresponding output signals, simulating the generation process of emotions and emotional memories in the human brain. This provides a hardware implementation solution for improving the human-computer interaction of bionic robots.

[0057] As shown in Figure 3, the thalamus module includes two similar circuits (representing the ability to process two signals simultaneously). One of the circuits includes two rectifier operational amplifiers, two transmission gate switches S1, S2 and an inverter. One of the rectifier operational amplifiers includes: operational amplifier A1-1, diode D1 and resistors R1-R3. The other rectifier operational amplifier includes: operational amplifier A1-2, diode D2 and resistors R4-R5; the Vre signal controls the transmission gate switches S1, S2 to select the signal Vin or Vre to input into the subsequent circuit. At the same time, if the Vin signal is selected, it will be processed by two rectifier operational amplifiers, and the Vin signal will be processed into a positive emotional signal VG1 and a negative emotional signal VB1.

[0058] Another circuit in the thalamus module includes operational amplifiers A1-3, A1-4, diodes D3, D4 and resistors R6-R10 and transmission gate switches S3, S4; the functions implemented are similar to the above, and are used to generate a positive emotional signal VG2 and a negative emotional signal VB2.

[0059] At the same time, the inputs Vin and Vre are controlled by Vc to input into the thalamus module. The thalamus module divides the Vin signal into a positive signal VG1 and a negative signal VB1, or selects the Vre signal to input into the sensory cortex module.

[0060] As shown Figure 4 in the figure, the somatosensory cortex module and the orbitofrontal cortex module include: a circuit for processing signals VG1, VG2, VForget1 and a circuit for processing signals VB1, VB2, VForget2;

[0061] The circuit for processing signals VG1, VG2, VForget1 includes: an adder, a differential proportional amplifier and a negative feedback integration loop; the adder includes operational amplifier A2-1 and resistors R10-R13, the differential proportional amplifier includes non-volatile memristor NVM1, operational amplifier A2-2 and resistor R14, and the negative feedback integration loop includes: operational amplifiers A2-3, A2-4, A2-5 and resistors R15-R19 and capacitor C1; the input signal from the thalamus module is integrated by the adder composed of operational amplifier A2-1 and resistors R10-R13, and the integrated emotion signal is input into the CA3 module, and at the same time, it is also input into the orbitofrontal module through the differential proportional amplifier; the orbitofrontal module receives the signal from the somatosensory cortex module, and is converted into a stable continuous signal VOG through the negative feedback integration loop composed of operational amplifiers A2-3, A2-4, A2-5 and resistors R15-R19 and capacitor C1, and is transmitted to the amygdala module;

[0062] The circuit for processing signals VB1, VB2, VForget2 includes: operational amplifiers A2-6, A2-7, A2-8, A2-9, A2-10 and resistors R20-R29 and capacitor C2; it is used to process the negative-polarity emotion signal and generate a continuous signal VOB to be transmitted to the amygdala module;

[0063] The output of the adder in the somatosensory cortex module is sent to the CA3 module, and the output of the orbitofrontal cortex module is sent to the amygdala module.

[0064] As shown Figure 5As shown, the amygdala module includes: three reverse adders, two comparators B3-1 and B3-2, two followers A3-3 and A3-5 to form a comparison module, and two POMS switch transistors. The first adder consists of operational amplifier A3-1, volatile memristor VM1, and resistors R30 - R31. The second adder consists of operational amplifier A3-4, volatile memristor VM2, and resistors R32 - R33. The third adder consists of two non-volatile memristors NVM3, NVM4, and resistor R34. The outputs of the first adder and the second adder serve as the two inputs of the third adder. The output terminal of the third adder serves as the input of the comparison module. The comparison module is composed of comparator B3-1 and follower A3-3, and comparator B3-2 and follower A3-5; the outputs of the two followers are respectively connected to the same PMOS switch transistor. The source of the PMOS switch transistor is connected to the follower, and the drain is connected to the CA1 module; the output of the amygdala module is controlled by VETM and input into the CA1 module; the amygdala module receives the emotional signals VOG and VOB processed by the orbitofrontal cortex module and simultaneously receives the long-term memories VLM+ and VLM- generated from the CA1 area. It receives the positive signal VOG and VLM+ through the first adder, and the negative signal VOB and VLM- through the second adder. The outputs of the first adder and the second adder are integrated by the third adder, and the signal integrated by the third adder generates emotional arousal signals EG1 and EB1 through the comparison module. The generated emotional arousal signals are transmitted to the CA1 module through two PMOS switches controlled by the VETM signal respectively.

[0065] As Figure 6 shown, the DG module includes: two rectifying operational amplifiers, two transmission gate switches S5, and an inverter. The first rectifying operational amplifier consists of operational amplifier A1-5, resistors R35 - R37, and diode D6. The second rectifying operational amplifier consists of operational amplifier A1-6, resistors R38 - R39, and diode D5; the control signal controls the transmission gate S5. When the transmission gate is turned on, the input signal Vin is preprocessed into a positive-polarity learning signal VL+ and a negative-polarity learning signal VL- through operational amplifiers A1-5, A1-6, resistors R35 - 39, and D5, D6.

[0066] Meanwhile, the input Vin is controlled by Vce and input into the CA3 module. The DG module divides the Vin signal into two signals VL+ and VL- of positive and negative polarities and inputs them into the CA3 module.

[0067] As Figure 7As shown, the CA3 module includes an emotion influence circuit, a learning signal circuit for processing positive signals, and a learning signal circuit for processing negative signals (the two learning signal circuits process the positive signal VL+ and the negative signal VL- respectively);

[0068] The emotion influence circuit includes: an inverting adder, a transmission gate switch S6, two comparators and two followers; the inverting adder includes: operational amplifier A4-1 and resistors R40-R42; comparator B4-2 and follower A4-5, and comparator B4-3 and follower A4-6 form a comparison module. The output of the adder in the sensory cortex module is sent to the CA3 module through the emotion influence circuit; when stimulated by the learning signal VL+ or VL-, the switch S6 conducts, and the VMCMG signal and VMCMB signal from the sensory cortex module will be integrated by the inverting adder and then input into the learning signal circuit through the switch S6. At the same time, the inverting adder is also connected to the comparison module, and by comparing the output of the inverting adder, the emotion valence signals EG2 and EB2 are obtained. The emotion valence signals EG2 and EB2 and the emotion arousal signals EG1 and EB1 constitute the emotion space representing emotions.

[0069] The learning signal circuit for processing positive signals includes: an inverting adder, an inverting proportional amplifier, and a comparison module composed of a comparator and a follower. The inverting adder includes: operational amplifier A4-2, volatile memristor VM3, and resistors R43-R44. The inverting proportional amplifier includes: operational amplifier A4-3, resistors R45 and R46. The comparison module includes: comparator B4-1 and follower A4-4. The processed positive-polarity learning signal VL+ from the DG module, through the inverting adder, first performs an addition operation with the signal from the emotion influence learning circuit. Then, through the comparison module, if the signal VSM+ is generated, it represents that the short-term memory has been generated at this time. Subsequently, the VSM+ signal is transmitted to the CA1 module, and long-term memory is generated in the CA1 module.

[0070] The learning signal circuit for processing negative signals includes: volatile memristor VM4, operational amplifiers A4-7, A4-8, A4-9, and resistors R47-R50; the functions implemented are similar to the above. It is used to process the negative-polarity learning signal VL-. Generate VSM- and input it into the subsequent CA1 module to generate long-term memory.

[0071] The output of the CA3 module is sent into the CA1 module.

[0072] As Figure 8 shown, the CA1 module includes: a circuit for processing the positive signal VSM+, a circuit for processing the negative signal VSM-, and a control circuit composed of four transistors;

[0073] Control circuit: It includes two pairs of NMOS and POMS transistors; when the control signal Vctr is at a high level, the control circuit will transmit a recall signal Vread through R52 and R56, and when Vctr is at a low level, the forgetting signal Vf will be transmitted through R52 and R56.

[0074] The circuit for processing the positive signal VSM+ includes: an inverting adder, an inverting amplifier, and a comparison module composed of a comparator and a follower; the inverting adder includes: operational amplifier A5-1, volatile memristors VM5, VM6, and resistors R51-R53; the inverting amplifier includes: operational amplifier A5-2, non-volatile memristor NVM5, and resistor R54; the comparison module includes comparator A5-3, follower A5-4, and a PMOS switch; the inverting adder receives the emotion arousal signals EG1, EB1 from the amygdala through VM5, VM6, receives the signal VSM+ from the CA3 module through resistor R51, receives the signal from the control circuit through R52, and the output of the inverting adder is input to the comparison module through the inverting proportional amplifier, and the generated signal LM+ representing long-term memory is transmitted to the amygdala module through the PMOS switch controlled by the control signal VMTE.

[0075] The circuit for processing the negative signal VSM- includes: an inverting adder, an inverting amplifier, and a comparison module composed of a comparator and a follower; the inverting adder includes: operational amplifier A5-5, volatile memristors VM7, VM8, and resistors R55-R57; the inverting amplifier includes: operational amplifier A5-6, non-volatile memristor NVM5, and resistor R58; the comparison module includes comparator A5-7, follower A5-8, and a PMOS switch; the functions implemented are similar to the above. It is used to receive the VSM- signal from the CA3 module, the EG1, EB1 signals from the amygdala, and the signal from the control circuit. And generate the signal LM- representing long-term memory and transmit it to the amygdala module through the PMOS switch controlled by VMTE.

[0076] The output of CA1 is controlled by the VMTE signal and input to the amygdala module.

[0077] Embodiment 2

[0078] In this embodiment, a simulation experiment is carried out on the emotion memory generation system in Embodiment 1;

[0079] 1. Simulate the emotion generation. The simulation process is as follows: First, apply an emotion signal VB1 with a value of -0.2V and a pulse width of 5ms to the circuit. After 0.2s, input the emotion signal VG1 and copy it to 0.1V with a pulse width of 10ms. The duration of this process is approximately 0.4s. After 0.5s, apply the emotion signal VG1 with a value of 0.2V and a pulse width of 10ms. After approximately 0.75s, apply the emotion signal VB1 with a value of -0.1V and a pulse width of 5ms. Continue until approximately 0.95s.

[0080] The simulation results are as Figure 9 shown. It can be seen from the figure that before 0.2s, only the negative emotion signal VB1 (0.2V, 5ms) exists, and the emotion arousal signal VEB1 is triggered at about 0.07s. After 0.2s, the positive emotion signal VG1 (0.1V, 10ms) is superimposed on the negative signal, resulting in the neutralization of the emotion signal and its ultimate disappearance for about 0.21s. This interaction generates an emotion valence signal EB2 with a duration of 5ms and an emotion arousal signal EB1 with a duration of about 0.14s, indicating the emergence of a sad emotion. After 0.21s, the absence of the emotion arousal signal VEB1 reflects the dynamics of emotion regulation. In the subsequent stage (from 0.5 seconds to 1 second), a similar pattern appears: the emotion valence signal EG2 lasts for 10 milliseconds, accompanied by the emotion arousal signal VEG1 lasting for about 0.22 seconds, indicating the generation of an excited emotion. This simulation effectively demonstrates the ability of the emotion regulation mechanism to maintain the stability and balance of the emotional state when dealing with various emotion signals.

[0081] 2. Simulate the impact of emotion on the learning rate. The simulation process is as follows: First, input the learning signals VL+ and VL- with an amplitude of 0.6V and a duration of 0.4S seconds. Observe the time when short-term memory is generated. Then, no excitation is applied at 1.1s to ensure that the short-term memory formed by the first excitation is forgotten. At 1.5s, apply the same stimulus as in the first stage, and at the same time input the emotion signal VG1 with a value of 0.3V and a pulse width of 10ms. Observe the time when short-term memory is generated at this time.

[0082] The simulation results are as Figure 10 shown. It can be seen from the figure that in the first stage (0 - 0.4s), the positive learning signal VL+ and the negative learning signal VL- with equal amplitudes but opposite polarities are applied simultaneously. In the second stage (1.4 - 1.8s), the same learning signals are used, and an additional emotion signal VG1 is introduced. The simulation results show that in the first stage, the learning rates of VL+ and VL- are comparable. However, in the second stage, the introduction of the VG1 signal leads to a significant increase in the learning rate of VL+ compared to VL-, thus proving an effective MCM. This simulation verifies the dual impact of positive and negative emotions on memory, indicating that the mismatch between the emotional state and information can enhance or hinder memory acquisition.

[0083] 3. Simulate the emotionally aroused memory. The simulation process is as follows: First, set the control signal VETM = 1V, and set the resistance values of the non-volatile memristors NVM5 and NVM6 to 11 kΩ to store long-term memories. Then, input the emotional signal VG1 for a duration of approximately 0.6 s, and observe the emotional arousal signal VEG1 generated by the emotional signal and the long-term memory VLM+ awakened by VEG1. After 0.7 s, input the emotional signal VB1 for a duration of approximately 0.6 s. Observe the emotional arousal signal VEB1 generated by the emotional signal and the long-term memory VLM- awakened by VEB1.

[0084] The simulation results are as Figure 11 shown. It can be seen from the figure that the emotional signal is divided into VG1 and VB1 through the thalamus module and is respectively input into the subsequent modules. When VETM = 1V, when a sufficiently strong emotion is aroused, it triggers the retrieval of memories related to that emotional state. Specifically, set the resistance values of the non-volatile memristors RNMV5 = RNMV6 = 1 kΩ to store long-term memories. In the time period from 0 to 0.7 s, input the positive emotional signal VG1, triggering an emotional arousal signal EG1 (strong enough to trigger memory retrieval) that lasts for approximately 0.6 s. At about 0.4 seconds, the long-term memory LM+ is successfully retrieved. Similarly, the negative signal VB triggers the emotional arousal signal EB1 after 0.7 s, thereby extracting the long-term memory LM-. When introducing positive or negative stimuli, a strong and persistent emotional arousal triggers the long-term memories resident in the hippocampus. Introducing positive or negative stimuli can both trigger strong and persistent emotional arousal, awakening the long-term memories stored in the hippocampus. This process is controlled by the VETM signal, effectively simulating the process of emotion-dependent memory (MDM).

[0085] 4. Simulate the influence of memory learning on emotional experience. The simulation process is as follows: First, set the control signal VMTE = 1V. Input the learning signal VL+ 1 s before and input the learning signal VL- 1 s later, and observe the short-term memories VSM+, VLM+ and VSM-, VLM- generated by the action of the learning signals, as well as the emotional arousal signals VEG1 and VEB1 caused by VLM+ and VLM-.

[0086] The simulation results are as Figure 12As shown, it can be seen from the figure that the learning signal is divided into VL+ and VL- by the DG module and input into the subsequent circuit. When VMTE = 1V, it indicates that the emotional experience is triggered by the formation of long-term memory. In the time range of 0 to 1 s, the appearance of the learning signal VL+ results in the formation time of short-term memory VSM+ being approximately 0.09 s. Subsequently, within approximately 0.1 s, the short-term memory is successfully consolidated into long-term memory VLM+. Under the regulation of VMTE, after the long-term memory is established, the emotional arousal signal EG1 appears approximately 0.4 s later and is accompanied by the formation of long-term memory. Similarly, within the time interval of 1.5 to 2 s, under the control of VMTE, the learning signal VL- can stimulate the emotional arousal signal EB1. During the process of learning and memory formation, when an individual comes into contact with learning information, emotional reactions often occur, which reflects the interaction between cognition and emotion. The above-mentioned simulation effectively mimics this process.

[0087] 5. For the simulation of robot applications, the robot obtains signals from the outside world through touch and vision. In Figure 13 (a) below, when the robot faces strong visual stimuli and slight tactile feedback, it can learn and recognize efficiently through the emotional memory circuit. This scenario cleverly mimics the process of the robot absorbing knowledge from a book, deepening the learning experience and cognitive understanding by touching the book. Due to the input of a high-level visual signal, the control signal VMTE is in a high-level state. During the learning process, the activation of VLM+ will trigger the generation of emotions. That is to say, when long-term memory LM+ is formed, the corresponding emotional experience will be stimulated. In another scenario, we simulated the actions of the robot in a dark environment. In this case, the tactile signal dominates and guides the robot to generate corresponding emotional reactions, such as panic or fear. These tactile-driven emotions not only give the robot rich emotional experiences but also help it make progress in memory and cognition. As Figure 13 (b) below shows, under the action of a strong tactile signal, the control signal VETM remains at a high level. During the exploration journey in the dark, the robot experiences negative emotion VEB1, and through this emotion, the formation of long-term memory LM- is further consolidated.

[0088] In the description of this specification, the descriptions referring to terms such as "one embodiment", "example", "specific example", etc. mean that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in a suitable manner in any one or more embodiments or examples.

[0089] The foregoing has shown and described the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments, and what is described in the above embodiments and the specification is only to illustrate the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements fall within the scope of the present invention claimed.

Claims

1. A memristor-based emotional memory generation system, characterized in that It includes a brain emotion learning model and a hippocampus model; the brain emotion learning model includes: a thalamus module, a sensory cortex module, an orbitofrontal cortex module, and an amygdala module; the hippocampus model includes: a DG module, a CA3 module, and a CA1 module; Under the stimulation of an emotion signal, the thalamus module first preprocesses the emotion signal, and then the sensory cortex module and the orbitofrontal module integrate and process the preprocessed emotion signal, and transmit the processed signal to the CA3 module and the amygdala module respectively to generate an output signal representing the emotion; Under the stimulation of a learning signal, the DG module first preprocesses the learning signal, and then the CA3 area is responsible for receiving the learning signal and generating an output signal representing short-term memory, which is transmitted to the CA1 area; after receiving the short-term memory signal, the CA1 area converts the short-term memory into long-term memory and stores it.

2. The emotional memory generation system based on a memristor according to claim 1, characterized in that, The preprocessing of the emotion signal by the thalamus module includes: receiving an external emotion signal, dividing the emotion signal into positive and negative polarity signals through the thalamus module, where the positive polarity signal represents a positive emotion signal and the negative polarity signal represents a negative emotion signal, and selecting an emotion signal or a forgetting signal to be transmitted to the sensory cortex module and the orbitofrontal module through a control signal.

3. The emotional memory generation system based on a memristor according to claim 1, wherein The integration and processing of the emotion signal by the sensory cortex module and the orbitofrontal module includes: the thalamus module can receive multiple emotion signals and preprocess the multiple emotion signals into multiple positive polarity signals and multiple negative polarity signals. The sensory cortex module first integrates the signals from the thalamus module and transmits the integrated signals to the CA3 module and the orbitofrontal module respectively. Subsequently, the orbitofrontal module extracts specific features from the sensory cortex module and evaluates their emotional significance.

4. The emotional memory generation system based on a memristor according to claim 1, wherein The thalamus module includes two circuits. One of the circuits includes: two rectifier operational amplifiers, two transmission gate switches S1, S2, and an inverter. The first rectifier operational amplifier includes: operational amplifier A1-1, resistors R1-R3, and diode D1. The second rectifier operational amplifier includes: operational amplifier A1-2, resistors R4-R5, and diode D2. The transmission gates S1, S2 are controlled by a control signal VC to select the Vin signal or Vre to be transmitted to the subsequent module. Among them, the Vin signal will be processed into positive and negative polarity signals by operational amplifiers A1-1, A1-2, resistors R1-R5, and two diodes D1, D2; Another circuit in the thalamus module includes operational amplifiers A1-3, A1-4, resistors R6-R9, and transmission gate switches S3, S4, and can also process two emotion signals Vin simultaneously.

5. The emotional memory generation system based on a memristor according to claim 4, wherein The sensory cortex module and the orbitofrontal cortex module include: a circuit for processing signals VG1, VG2, VForget1 and a circuit for processing signals VB1, VB2, VForget2; The circuit for processing signals VG1, VG2, and VForget1 includes: an adder, a differential proportional amplifier, and a negative feedback integration loop; the adder includes operational amplifier A2-1 and resistors R10-R13, the differential proportional amplifier includes non-volatile memristor NVM1, operational amplifier A2-2, and resistor R14, and the negative feedback integration loop includes operational amplifiers A2-3, A2-4, A2-5, resistors R15-R19, and capacitor C1; the input signal from the thalamus module is integrated by the adder composed of operational amplifier A2-1 and resistors R10-R13, and the integrated emotion signal is input into the CA3 module. At the same time, it is also input into the orbitofrontal module through the differential proportional amplifier; the orbitofrontal module receives signals from the sensory cortex module and is transformed into a stable continuous signal VOG through the negative feedback integration loop composed of operational amplifiers A2-3, A2-4, A2-5, resistors R15-R19, and capacitor C1, and is transmitted to the amygdala module; The circuit for processing signals VB1, VB2, and VForget2 includes operational amplifiers A2-6, A2-7, A2-8, A2-9, A2-10, resistors R20-R29, and capacitor C2; it is used to process negative-polarity emotion signals and generate a continuous signal VOB for transmission to the amygdala module; The output of the adder in the sensory cortex module is fed into the CA3 module, and the output of the orbitofrontal cortex module is fed into the amygdala module.

6. The emotional memory generation system based on a memristor according to claim 5, wherein The amygdala module includes: three inverting adders, two comparators B3-1 and B3-2, two followers A3-3, A3-5, and two PMOS switch transistors; the first adder includes operational amplifier A3-1, resistors R30, R31, and volatile memristor VM1, the second adder includes operational amplifier A3-4, resistors R32, R33, and volatile memristor VM2; the third adder includes operational amplifier A3-2, non-volatile memristors NVM3, NVM4, and resistor R34; the first adder and the second adder receive signals from the orbitofrontal cortex module and the CA1 module, and the third adder receives the outputs of the first adder and the second adder; the two comparators B3-1 and B3-2 are respectively connected to the followers A3-3 and A3-5; the outputs of the two followers are respectively connected to the same PMOS switch transistor; the source of the PMOS switch transistor is connected to the follower, and the drain is connected to the CA1 module; the output of the amygdala module is controlled by VETM and input into the CA1 module; The amygdala module receives the emotional signals VOG and VOB processed by the orbitofrontal cortex module. VOG and VOB simultaneously receive the long-term memories VLM+ and VLM- generated from the CA1 region. The first adder receives the positive-polarity signal VOG and VLM+, and the second adder receives the negative-polarity signal VOB and VLM-. The outputs of the first adder and the second adder are integrated by the third adder, and the signal integrated by the third adder generates the emotional arousal signals EG1 and EB1 through the comparison module. The generated emotional arousal signals are transmitted to the CA1 module through two PMOS switches controlled by the VETM signal respectively.

7. The emotional memory generation system based on a memristor according to claim 6, wherein The DG module includes: a rectifying operational amplifier, a transmission gate switch S5, and an inverter; the rectifying operational amplifier includes: operational amplifiers A1-5, A1-6, resistors R35-39, and diodes D5 and D6; the control signal controls the transmission gate S5. When the transmission gate is turned on, the input signal Vin is preprocessed into a positive-polarity learning signal VL+ and a negative-polarity learning signal VL- through the operational amplifiers A1-5, A1-6, resistors R35-39, and D5, D6. The DG module divides the Vin signal into two signals VL+ and VL- of positive and negative polarities and inputs them into the CA3 module.

8. The emotional memory generation system based on a memristor according to claim 7, wherein The CA3 module includes: an emotion influence circuit, a learning signal circuit for processing positive signals, and a learning signal circuit for processing negative signals. The emotion influence circuit includes: a reverse adder, a transmission gate switch S6, two comparators, and two followers; the reverse adder includes: operational amplifier A4-1 and resistors R40-R42; the comparator B4-2 and the follower A4-5, and the comparator B4-3 and the follower A4-6 form a comparison module. The output of the adder in the sensory cortex module is sent into the CA3 module through the emotion influence circuit; when stimulated by the learning signals VL+ or VL-, the VMCMG signal and the VMCMB signal from the sensory cortex module are integrated by the reverse adder and then input into the learning signal circuit; at the same time, the reverse adder is also connected to the comparison module. By comparing the output of the reverse adder, the emotional valence signals EG2 and EB2 are obtained; the emotional valence signals EG2 and EB2 and the emotional arousal signals EG1 and EB1 constitute the emotional space representing emotions. The learning signal circuit for processing positive signals includes: a reverse adder, a reverse proportional amplifier, and a comparison module composed of a comparator and a follower. The reverse adder includes: operational amplifier A4-2, a volatile memristor VM3, and resistors R43-R44. The reverse proportional amplifier includes: operational amplifier A4-3, resistors R45 and R46. The comparison module includes: comparator B4-1 and follower A4-4. The positive learning signal VL+ processed by the DG module, through a reverse adder, first performs an addition operation with the signal from the emotion-influenced learning circuit; then, through the comparison module, if the signal VSM+ is generated, it represents that the short-term memory has been generated at this time; subsequently, the VSM+ signal is transmitted to the CA1 module, and long-term memory is generated in the CA1 module. The learning signal circuit for processing negative signals includes: a volatile memristor VM4, operational amplifiers A4-7, A4-8, A4-9, and resistors R47-R50, which are used to process the negative-polarity learning signal VL-, generate VSM-, and input it into the subsequent CA1 module to generate long-term memory.

9. The mood memory generation system based on a memristor according to claim 8, characterized in that, The CA1 module includes: a circuit for processing the positive signal VSM+, a circuit for processing the negative signal VSM-, and a control circuit composed of four transistors. Control circuit: includes two pairs of NMOS and POMS transistors; when the control signal Vctr is at a high level, the control circuit will transmit a recall signal Vread through R52 and R56, and when Vctr is at a low level, the forgetting signal Vf will be transmitted through R52 and R56. The circuit for processing the positive signal VSM+ includes: a reverse adder, a reverse amplifier, and a comparison module composed of a comparator and a follower; the reverse adder includes: operational amplifier A5-1, volatile memristors VM5, VM6, and resistors R51-R53; the reverse amplifier includes: operational amplifier A5-2, non-volatile memristor NVM5, and resistor R54; the comparison module includes comparator A5-3, follower A5-4, and a PMOS switch. The reverse adder receives the emotion arousal signals EG1, EB1 from the amygdala through VM5, VM6, receives the signal VSM+ from the CA3 module through resistor R51, and receives the signal from the control circuit through R52. The output of the reverse adder is input into the comparison module through a reverse proportional amplifier, and the generated signal LM+ representing long-term memory is transmitted to the amygdala module through the PMOS switch controlled by the control signal VMTE. The circuit for processing the negative signal VSM- includes: a reverse adder, a reverse amplifier, and a comparison module composed of a comparator and a follower; the reverse adder includes: operational amplifier A5-5, volatile memristors VM7, VM8, and resistors R55-R57; the reverse amplifier includes: operational amplifier A5-6, non-volatile memristor NVM5, and resistor R58; the comparison module includes comparator A5-7, follower A5-8, and a PMOS switch; it is used to receive the VSM- signal from the CA3 module, the EG1, EB1 signals from the amygdala, and the signal from the control circuit, and generate the signal LM- representing long-term memory, which is transmitted to the amygdala module through the PMOS switch controlled by VMTE. The output of CA1 is controlled by the VMTE signal and input into the amygdala module.

10. A robot, characterized in that, Including a memristor-based emotion memory generation system according to any one of claims 1-9.

Images