CN-121996076-A - Multi-physical-quantity cooperative sensing brain-like synaptic device and dynamic environment self-adaption method thereof

CN121996076ACN 121996076 ACN121996076 ACN 121996076ACN-121996076-A

Abstract

The invention discloses a multi-physical quantity collaborative sensing brain-like synaptic device and a dynamic environment self-adapting method thereof, relating to the technical field of brain-like synaptic devices and multi-physical quantity sensing, comprising a multi-physical quantity sensing layer, a temperature sensing unit, a humidity sensing unit and a gas sensing unit; the memristor synaptic modulation device comprises a memristor synaptic modulation layer, a cooperative control unit, a read-write control module, an interface and an integration module, wherein the structures are integrated on a wearable carrier to realize hardware level cooperation of multi-physical quantity sensing and memristor synaptic weight regulation. According to the invention, through dynamically updated interference source characteristics and combining with the linkage of a lightweight machine learning classifier, the accurate identification of the interference type and the differential anti-interference processing are realized, the defects that in the prior art, a filtering threshold is fixed, a dynamic interference feature library is not available, and multiple interferences in a complex environment are difficult to deal with are overcome, and meanwhile, the clutter misjudgment probability is reduced through dynamically adjusting the feature matching threshold and the lowest threshold.

Inventors

ZHAO CHUN
ZHAO ZISHEN

Assignees

深圳市华芯邦科技有限公司

Dates

Publication Date: 20260508
Application Date: 20260410

Claims (10)

1. A multi-physical quantity cooperative sensing brain-like synaptic device comprising: the multi-physical quantity sensing layer comprises a temperature sensing unit, a humidity sensing unit and a gas sensing unit and is used for collecting temperature, humidity and gas concentration signals in the environment; the memristive synaptic modulation layer adopts a memristive structure modulated by reversible conductance and is used for simulating synaptic weights; The cooperative control unit is respectively and electrically connected with the multi-physical-quantity sensing layer and the memristive synaptic modulation layer and is used for generating a weight regulation instruction according to signals acquired by the multi-physical-quantity sensing layer and the current weight state of the memristive synaptic modulation layer; The read-write control module is electrically connected with the cooperative control unit and the memristive synaptic modulation layer and is used for regulating and controlling the weight of the memristive synaptic modulation layer according to the weight regulating and controlling instruction; The interface and the integrated module are electrically connected with the cooperative control unit and the read-write control module and are used for communicating with an external system; the structure is integrated on the wearable carrier, so that hardware-level cooperation of multi-physical quantity sensing and memristor synaptic weight regulation is realized.
2. The dynamic environment self-adaptive method for multi-physical-quantity cooperative sensing is characterized in that the self-adaptive method comprises the multi-physical-quantity cooperative sensing brain-like synaptic device as claimed in claim 1, and specifically comprises the following steps: S1, acquiring three types of physical quantity signals of temperature, humidity and gas concentration in an environment in real time through a temperature-sensitive unit, a humidity-sensitive unit and a gas-sensitive unit of a multi-physical quantity sensing layer; S2, the multi-physical quantity sensing layer carries out preliminary filtering on the acquired multi-path physical quantity signals, and after obvious impurities are removed, the signals are transmitted to the cooperative control unit; s3, the read-write control module reads the current weight state of the memristor synaptic modulation layer in real time and transmits the weight state to the cooperative control unit; s4, the cooperative control unit performs feature extraction and analysis on the received physical quantity signals, and generates a targeted weight regulation instruction by combining the weight state of the memristor synaptic modulation layer and transmits the targeted weight regulation instruction to the read-write control module; s5, the read-write control module dynamically adjusts pulse parameters according to the regulation and control instruction, and carries out self-adaptive regulation and control on the weight of the memristor synaptic modulation layer so as to realize accurate matching of the physical quantity signal and the synaptic weight; And S6, detecting the regulation and control effect in real time by the cooperative control unit, comparing memristor weight states before and after regulation and control with the accuracy of the sensing signal, and dynamically optimizing the regulation and control logic to form closed-loop self-adaptive regulation and control.
3. The dynamic environment adaptive method for multi-physical-quantity cooperative sensing according to claim 2, wherein the physical-quantity signal in S1 specifically comprises: the temperature sensitive unit, the humidity sensitive unit and the gas sensitive unit of the multi-physical-quantity sensing layer are in partition collection configuration, the inner side unit is attached to the skin of a human body to collect the related physical quantity of the human body, and the outer side unit is exposed to the environment to collect the environmental physical quantity; According to the change characteristics of the environmental physical quantity, the acquisition sensitivity of each unit is dynamically adjusted; And after the acquired multipath physical quantity signals ensure the signal integrity, transmitting the acquired multipath physical quantity signals to a subsequent processing link.
4. The dynamic environment adaptive method of multi-physical-quantity cooperative sensing according to claim 2, wherein the preliminary filtering and transmission in S2 specifically includes: The method comprises the steps of removing high-frequency interference by adopting a differential filtering mode aiming at signals of different physical quantities, and synchronously transmitting an original acquisition signal and a filtered signal to a cooperative control unit; The cooperative control unit acquires an original signal before filtering through a sliding window, extracts signal statistical characteristics and a noise substrate, dynamically calculates clutter rejection threshold values based on signal distribution characteristics, and adaptively adjusts threshold value parameters; establishing a joint threshold judgment mechanism by combining multiple physical quantity coupling characteristics, introducing a lightweight machine learning classifier to identify signal mutation attributes, and dynamically adjusting threshold parameters to distinguish environmental changes from interference; Recording a signal misjudgment case, optimizing threshold calculation parameters through a closed loop feedback mechanism, ageing an adapter piece and changing the environment, and supporting periodic threshold recalibration; the event-driven mode is adopted to realize the updating of the threshold value; removing abnormal clutter according to the dynamically adjusted threshold value, and reserving effective physical quantity signals; And (3) carrying out time sequence synchronization alignment on the filtered multipath physical quantity signals, and transmitting the signals to the cooperative control unit after ensuring consistent transmission time sequence.
5. The method for dynamic environment adaptation based on multi-physical-quantity cooperative sensing according to claim 2, wherein the controlling and transmitting of the current weight state in S3 specifically comprises: The read-write control module sends detection pulses to the memristor synaptic modulation layer in a pulse detection mode to acquire memristor conductance response signals; according to the memristor conductivity response signal, calculating a current memristor synaptic weight value through a preset algorithm, defining a weight stable range, and synchronously recording a weight drift amount; And packaging the current weight value, the weight stabilizing range and the drift amount, and transmitting the current weight value, the weight stabilizing range and the drift amount to the cooperative control unit through a preset bus.
6. The method for dynamic environment adaptation based on multi-physical-quantity cooperative sensing according to claim 2, wherein the generating a targeted weight regulation instruction in S4 specifically includes: the cooperative control unit performs feature extraction on the received multipath physical quantity signals, and extracts three core features of the change rate, the fluctuation amplitude and the steady-state duration of the signals; According to the extracted signal characteristics, judging the influence degree of each physical quantity on memristive synaptic weights, and dynamically distributing the regulation and control priority of each physical quantity in combination with the application requirements of wearable scenes; judging whether the current weight deviates from the adaptive range or not by combining with the memristor weight state transmitted by the read-write control module, and generating a targeted weight regulation instruction comprising a pulse amplitude, a pulse width and a regulation step length; and encoding the regulation and control instruction and transmitting the regulation and control instruction to a read-write control module.
7. The method for adaptive sensing of dynamic environment by cooperative sensing of multiple physical quantities according to claim 2, wherein the adaptive adjustment and control of the weight in S5 specifically comprises: the read-write control module analyzes the received regulation and control instruction, extracts core regulation and control parameters and determines a regulation and control mode; the output pulse parameters are dynamically regulated according to the regulation and control mode, the judgment of the weight deviation degree adopts two modes of quantization or dynamic threshold, and small-step fine adjustment or large-step rapid regulation and control are respectively adopted according to the deviation degree; outputting an adjusted pulse signal to the memristor synaptic modulation layer, monitoring weight change in real time, and guaranteeing regulation and control accuracy; In the regulation and control process, the corresponding relation between pulse parameters and weight change is synchronously recorded, and data support is provided for subsequent regulation and control optimization.
8. The dynamic environment self-adaption method of multi-physical quantity collaborative awareness according to claim 2, wherein the adjusting logic in S6 specifically comprises: The cooperative control unit receives the memristor weight state after regulation and control fed back by the read-write control module, compares the weight state before regulation and control, calculates weight adjustment deviation, and judges whether the regulation and control effect reaches the standard; synchronously acquiring physical quantity sensing signals after regulation and control, comparing signal precision before regulation and control, and analyzing influence of interference factors on regulation and control effects; if the regulation effect does not reach the standard, optimizing the regulation pulse parameters and regulating the regulation logic according to the weight regulation deviation and the signal precision change; And storing the optimized regulation logic to form a closed loop iteration mechanism, so as to ensure that the subsequent regulation effect is continuously improved.
9. The method for adaptive sensing of dynamic environment with multiple physical quantities according to claim 2, wherein the acquisition process in S1 further comprises adaptive adjustment of sampling timing, specifically comprising: The cooperative control unit detects the change rate of each physical quantity in real time and divides the slow rate and the fast rate; According to the change rate level of the physical quantity, the sampling time sequence of the sensing layer is dynamically adjusted, the sampling interval is prolonged at a slow rate, and the sampling interval is shortened at a fast rate; the cooperative control unit synchronously links the regulation and control time sequence of the read-write control module, ensures that sampling and regulation of the sensing layer are synchronous, and avoids signal dislocation.
10. The method for dynamic environment adaptation based on multi-physical-quantity cooperative sensing according to claim 2, wherein the preliminary filtering process in S2 further comprises: presetting an interference source feature library and storing a historical interference sample based on the characteristics of each unit of the multi-physical quantity sensing layer, and storing the interference source feature library and the historical interference sample in a cooperative control unit; The cooperative control unit extracts the signal characteristics before filtering, calculates the similarity between the signal characteristics and the historical interference samples in the characteristic library, and dynamically adjusts the characteristic matching threshold; if the similarity between the signal and all the historical interference samples is lower than a preset minimum threshold, judging that the signal is a suspected new interference type, automatically marking and recording the characteristics of the suspected new interference type, and finishing the automatic discovery of the new interference; The cooperative control unit shares the signal characteristics and the similarity result to a lightweight machine learning classifier, multiplexes the classification result to verify the interference type, executes a differential anti-interference strategy according to the interference source type, and eliminates abnormal signals; And after the anti-interference processing is finished, the subsequent filtering and signal transmission are performed, so that the signal purity is ensured, meanwhile, the newly found interference characteristics are updated periodically, the matching threshold value is optimized, and the updating period can be adapted to a scene as required.

Description

Multi-physical-quantity cooperative sensing brain-like synaptic device and dynamic environment self-adaption method thereof Technical Field The invention relates to the technical field of brain-like synapse devices and multi-physical-quantity sensing, in particular to a brain-like synapse device and a dynamic environment self-adaption method thereof. Background Along with the rapid development of wearable electronic equipment and brain-like computing technology, the real-time sensing and accurate regulation and control requirements on multiple physical quantities (such as temperature, humidity and gas concentration) are increasingly urgent, and particularly in wearable application scenes such as health monitoring and environment sensing, devices are required to have high sensitivity, low power consumption and dynamic environment adaptation capability, the weight regulation and control characteristics of brain-like synapses can be simulated, the cooperative optimization of sensing signals and regulation and control logic is realized, and the efficient interference recognition and anti-interference capability and the closed loop iterative optimization capability of the regulation and control logic are required to be provided so as to cope with multiple interferences in complex dynamic environments and ensure the sensing and regulation and control precision. At present, the existing multi-physical-quantity sensing devices are mainly designed by adopting single physical-quantity sensing units, the sensing units lack of cooperative linkage, synchronous acquisition and coupling analysis of multi-physical-quantity signals cannot be realized, the acquisition sensitivity is fixed, and the dynamic-change environment scene is difficult to adapt. In the aspect of self-adaptive regulation of dynamic environment, the filtering processing of signals with multiple physical quantities adopts a fixed threshold mode, the threshold parameters cannot be dynamically regulated according to the signal distribution characteristics and the environmental interference change, the problems of clutter misjudgment, effective signal loss and the like are easy to occur, a dynamically updated interference source feature library is not established, the linkage with a machine learning classifier is lacking, the precise identification and the differential anti-interference processing of interference types cannot be realized, the treatment of multiple interferences in a complex environment is difficult to deal with, the regulation of memristive synaptic weights adopts fixed logic, the self-adaptive response to the environmental physical quantity change is lacking, the regulation strategy cannot be dynamically regulated according to the sensing signal features, a closed-loop optimization mechanism is not available after the regulation, the regulation logic cannot be reversely optimized according to the regulation effect, and the stability of the regulation precision is difficult to ensure. Disclosure of Invention The invention aims to provide a multi-physical-quantity collaborative sensing brain-like synaptic device and a dynamic environment self-adapting method thereof, so as to solve the problems in the background art. In order to achieve the above purpose, the invention provides the following technical scheme that the multi-physical-quantity cooperative sensing brain-like synaptic device comprises: the multi-physical quantity sensing layer comprises a temperature sensing unit, a humidity sensing unit and a gas sensing unit and is used for collecting temperature, humidity and gas concentration signals in the environment; the memristive synaptic modulation layer adopts a memristive structure modulated by reversible conductance and is used for simulating synaptic weights; The cooperative control unit is respectively and electrically connected with the multi-physical-quantity sensing layer and the memristive synaptic modulation layer and is used for generating a weight regulation instruction according to signals acquired by the multi-physical-quantity sensing layer and the current weight state of the memristive synaptic modulation layer; The read-write control module is electrically connected with the cooperative control unit and the memristive synaptic modulation layer and is used for regulating and controlling the weight of the memristive synaptic modulation layer according to the weight regulating and controlling instruction; The interface and the integrated module are electrically connected with the cooperative control unit and the read-write control module and are used for communicating with an external system; the structure is integrated on the wearable carrier, so that hardware-level cooperation of multi-physical quantity sensing and memristor synaptic weight regulation is realized. The dynamic environment self-adaption method for multi-physical quantity collaborative sensing specifically comprises the following steps: S1, acquiring three