Sensor array information detection system and design method thereof

Abstract

The invention provides a sensor array information detection system and a design method thereof. The system comprises a hardware circuit board taking a microprocessor as a core, and a sensor array, a power supply, a power supply management unit and a core microprocessor circuit which are arranged on the hardware circuit board, the power supply is electrically connected with the core microprocessor circuit and the sensor array through the power supply management unit; the power supply management unit provides bias voltage for each sensor forming the sensor array; the embedded application program is installed in the microprocessor, the application program is installed in a mobile phone or a computer and corresponds to the embedded application program, the computer software program is used for controlling the whole sensor array information detection system to work, and a plurality of function modules for filtering are arranged in the embedded application program; and the user mobile phone or the computer is in communication connection with the microprocessor. The system has the advantages of high noise interference resistance, low power consumption, universality, small size, light weight and the like.

Description

Technical Field

[0001] The present invention belongs to the technical field of sensors, and particularly relates to a sensor array information detection system and a design method thereof. Background Art

[0002] A sensor array - a sensor combination formed by arranging multiple sensors in an array form. Due to the mutual cooperation among multiple sensors, compared with a single sensor, it can improve the ability to sense and process information (signals). In particular, in the case where some sensors in the sensor array fail, as long as other sensors can still work properly, the entire information detection system including the sensor array can still effectively sense and process information, thus ensuring the normal operation of the entire detection system (in this case, the performance of the entire detection system will also decrease to some extent). The above characteristics make the sensor array significantly superior to a single sensor in terms of performance and reliability, and thus it is widely used in many fields such as industrial production, environmental detection, and healthcare, and has attracted the high attention of many researchers in this field and related technical fields.

[0003] In recent years, relevant researchers have conducted relatively in-depth research on sensor arrays and the entire information detection system including sensor arrays, and have achieved many research results. Some of these research results have been made public in the form of patents or patent applications. For example, the invention patent application with the publication number CN119197855A discloses a matrix circuit applicable to a sensor array and a sensor array, and the matrix circuit includes:

[0004] Multiple sensors, each sensor includes multiple sensor components, and each sensor component is arranged at different positions on the sensor body of the sensor for collecting sensing information at that position. Among them, each sensor component includes a set of connection terminals;

[0005] A wiring module, including multiple row lines and multiple column lines. A set of row lines and a set of column lines form a first-level matrix in the matrix circuit, and each first-level matrix is arranged in a matrix to form a second-level matrix in the matrix circuit. Among them, a set of row lines includes at least one row line, and a set of column lines includes at least one column line.

[0006] Among them, each sensor is correspondingly arranged with each first-level matrix, and the connection terminals of each sensor component of each sensor are connected to the row lines and column lines of the corresponding first-level matrix.

[0007] Based on the above technical solutions, the invention patent application No. CN119197855A further discloses additional technical means such as a calculation module (including an array signal acquisition circuit and a calculation unit), and on this basis, discloses a sensor array including the matrix circuit (in fact, discloses the entire signal detection system including the sensor array).

[0008] For another example, the invention patent with the authorization announcement No. CN 112730527 B discloses a gas detection system based on a MEMS gas sensor array, including:

[0009] A MEMS gas sensor array, a time-division multiplexed multi-channel resistance-frequency conversion circuit, a programmable heater circuit, an EEPROM, a tunable on-chip oscillator, a power-on self-reset circuit, and a digital control circuit; the MEMS gas sensor array includes a plurality of gas sensors; the gas sensor includes a heater resistance and a gas-sensitive material resistance;

[0010] The MEMS gas sensor array, the multi-channel resistance-frequency conversion circuit, the heater circuit, the EEPROM, the on-chip oscillator, and the power-on self-reset circuit are respectively connected to the corresponding pins of the digital control circuit;

[0011] Among them, the MEMS gas sensor array is used to convert the gas information in the environment into the change of the gas-sensitive material resistance; the multi-channel resistance-frequency conversion circuit is used to convert the resistance value of the gas-sensitive material in the selected channel into a square wave signal with a corresponding frequency; the heater circuit is used to configure the heating voltage of the heater resistance; the EEPROM is used to store system configuration parameters and user data; the on-chip oscillator is used to generate a stable clock signal required by the system; the digital control circuit is used to complete the functions of controlling the working mode of the on-chip circuit, frequency measurement, data storage, and I2C communication;

[0012] The multi-channel resistance-frequency conversion circuit includes: a resistance conversion current circuit, which is used to drive a reference voltage to both ends of the gas-sensitive material resistance to form a detection current related to the gas-sensitive material resistance; a CMOS current mirror circuit, which is used to reduce the detection current to form an image current; an integration circuit, which is used to output a periodic triangular wave signal based on the image current; a bistable hysteresis comparator circuit, which is used to output a periodic square wave signal based on the triangular wave signal;

[0013] The bistable hysteresis comparator circuit includes: a second high-gain operational amplifier, a third high-gain operational amplifier, a first resistor, and a second resistor; the non-inverting input terminal of the second high-gain operational amplifier is connected to the reference voltage, and its inverting input terminal is connected to its output terminal; the inverting input terminal of the third high-gain operational amplifier is connected to the fourth node, its non-inverting input terminal is connected to its output terminal through the first resistor, and is connected to the output terminal of the second high-gain operational amplifier through the second resistor; wherein, the fourth node is connected to the CMOS current mirror circuit; the output terminal of the third high-gain operational amplifier is connected with a plurality of cascaded inverters.

[0014] In the above gas detection system, the CMOS current mirror circuit includes: a PMOS current mirror and an NMOS current mirror;

[0015] The PMOS current mirror includes: a first transistor to a sixth transistor; the gates of the first transistor, the second transistor, and the fifth transistor are connected, and their sources are respectively connected to the power supply terminal through separate digitally controlled switches. The drain of the first transistor is connected to its own gate and the source of the third transistor. The drain of the second transistor is connected to the source of the fourth transistor. The drain of the fifth transistor is connected to the source of the sixth transistor. The gates of the third transistor, the fourth transistor, and the sixth transistor are connected. The drain of the third transistor is connected to the second node, and the second node is connected to the resistance conversion current circuit. The drain of the fourth transistor is connected to the third node, and the drain of the sixth transistor is connected to the fourth node;

[0016] The NMOS current mirror includes: a seventh transistor to a tenth transistor; the gates of the seventh transistor and the eighth transistor are connected. The drain of the seventh transistor is connected to the third node, and the drain of the eighth transistor is connected to the fourth node. The source of the seventh transistor is connected to the gate and the drain of the ninth transistor. The source of the eighth transistor is connected to the drain of the tenth transistor. The sources of the ninth transistor and the tenth transistor are respectively grounded through separate digitally controlled switches.

[0017] Based on the above technical solutions, the invention patent of CN112730527B also discloses a plurality of additional technical means.

[0018] There are many patents or patent applications related to sensor arrays and the information detection systems they consist of, which are not listed here one by one. In general, the technical solutions disclosed in the relevant patents or patent applications have promoted the advancement of sensor technology to varying degrees. For example, the technical solution disclosed in the invention patent CN112730527B reduces the operating voltage and circuit power consumption, improves the selectivity of the sensor, reduces the cross-sensitivity of the sensor, and improves the detection range and detection accuracy of the resistance of the gas-sensitive material in the gas sensor, which facilitates sensor data collection and processing. For another example, the technical solution disclosed in the invention patent application No. CN119197855A solves the problem of flexible wiring when the sensor forms an array, and has high scalability, can reduce the number of leads, thereby saving costs and optimizing the layout.

[0019] However, the technical solutions disclosed in the existing relevant patents or patent applications, as well as the relevant technical solutions disclosed in other ways (such as journal articles, master's and doctoral theses or public use), also have some technical defects, which are mainly manifested in the following three aspects:

[0020] (1) Noise: When the existing sensor array information detection system works in a harsh environment, it will be interfered by the noise in the environment, resulting in reduced sensitivity, accuracy and stability of the sensors in the sensor array information detection system, affecting the performance of the system.

[0021] (2) Power consumption: The sensors in the sensor array information detection system need to operate continuously under low power conditions. Therefore, how to optimize circuit design and energy management to achieve efficient energy utilization of the entire system and extend its service life, especially extending the service life of the sensor, is one of the key technical issues to be solved in this field. Although the technical solution disclosed in the invention patent application No. CN119197855A is indeed desirable in optimizing circuit design, it still has the technical defect of high power consumption.

[0022] (3) Universality: Existing sensor arrays that have been put into practical use and the information detection systems they consist of are all subject to specific environmental conditions (such as temperature, humidity, pressure, etc.), making it difficult for them to adapt to different environmental conditions, especially to maintain stable operation under extreme conditions. Therefore, they have the technical defect of weak universality.

[0023] In addition to the above technical defects, the hardware circuit boards in the existing sensor array information detection system are generally too large, which is not conducive to the design of portable devices and increases the burden on users. The oversized size also means that the weight of the entire system increases, which can easily cause fatigue to users during use, especially in mobile and wearable devices. In addition, the oversized size will also increase the production cost and energy consumption of the product, which is not conducive to the promotion and popularization of the product. Summary of the Invention

[0024] The object of the present invention is to provide a sensor array information detection system with strong anti-noise interference ability, low power consumption, universality, small volume and light weight, so as to overcome the defects of the above-mentioned prior art.

[0025] To achieve the above object, the present invention adopts the following technical solutions:

[0026] A sensor array information detection system includes a hardware circuit board with a microprocessor as the core, and a sensor array, a power supply, a power management unit, and a core microprocessor circuit installed on the hardware circuit board; the power supply is electrically connected to the core microprocessor circuit and the sensor array through the power management unit; the power management unit provides a bias voltage for each sensor constituting the sensor array; the sensor array information detection system further includes an embedded application program installed in the microprocessor, an application program corresponding to the embedded application program installed in a mobile phone or a computer, and a computer software program for controlling the operation of the entire sensor array information detection system. The embedded application program is provided with a median filtering function module, a first-order lag filtering function module, a Savitzky-Golay smoothing filtering function module, and a Gaussian filtering function module; the user's mobile phone or computer is communicatively connected to the microprocessor.

[0027] On the basis of the above technical solutions, the present invention can be supplemented with the following technical means to better or more pertinently solve the technical problems to be solved by the present invention:

[0028] The sensor array is a flexible sensor array.

[0029] Further, the sensor array is a strain sensor array.

[0030] Further, the sensor array is a gas sensor array.

[0031] Preferably, the gas sensor array is a MEMS gas sensor array.

[0032] Further, the computer software program is further provided with a Savitzky-Golay smoothing filtering function module (Savitzky-Golay smoothing filter).

[0033] Further, the computer software program is further provided with a Gaussian filtering function module (Gaussian filter)

[0034] The present invention further provides a design method for a sensor array information detection system, including the following steps:

[0035] Step 1, design the overall architecture: including the design of a sensor array, a core microprocessor circuit, an embedded application installed in the microprocessor, an application corresponding to the embedded application installed in a mobile phone or a computer, and a computer software program for controlling the operation of the entire sensor array information detection system;

[0036] Step 2, design the core circuit board: select a hardware circuit board with a microprocessor as the core, and transmit and receive the sampled data via Bluetooth through digital filtering and data encapsulation;

[0037] Step 3, design the algorithm: select the KNN algorithm and the baseline calibration algorithm;

[0038] Step 4, design the signal analysis: add a threshold determination mechanism to the KNN algorithm, that is, the T-KNN method, compare the data with a predetermined threshold, and eliminate signals below the threshold;

[0039] Step 5, optimize the performance: select a suitable time window and calculate the average value of the data within it.

[0040] Furthermore, a gesture recognition algorithm process is set in the computer software program, and the gesture recognition algorithm process includes the following steps:

[0041] S1, discriminate the original data set: after obtaining the original gesture action data from the lower computer, first discriminate the data. If the small threshold ≤ data ≤ large threshold, the data is determined to be valid. If the data < small threshold or the data > large threshold, the data is determined to be invalid;

[0042] S2, initialize the gesture sample data set: mark the data after discrimination, and divide the total samples into a training sample set and a test sample set, and label them according to the data of different people in the collected samples and different actions;

[0043] S3, calculate the Euclidean distance: that is, the distance of the straight line between two points. By calculating the true distance between two points in the m-dimensional space or the natural length of the vector (that is, the distance from this point to the origin), the nearest neighbor of a given sample can be identified. The specific calculation formula is as follows:

[0044]

[0045] In the formula, d is the distance between two points U and V, x iu and x iv are m features on the sample data;

[0046] S4, sort from small to large: according to the calculated Euclidean distance, sort them from small to large, and each data can be mapped to a label;

[0047] S5. Obtain the labels of the top K data: According to the nearest neighbor idea, circle the K objects with the closest distance as the neighbors of the test object. Determine the final sample category through majority voting;

[0048] S6. Obtain the result of gesture recognition: Match and connect with the host computer software interface, and output the result of majority voting.

[0049] Compared with the prior art, the main beneficial effects of the present invention are as follows:

[0050] First, reduce power consumption: The present invention introduces a variety of power consumption control strategies, and by introducing multiple bias voltages and a low-power self-wake-up mechanism, significantly reduces the energy consumption of the sensor system, extends the battery life, and is particularly suitable for application scenarios with long-term operation.

[0051] Second, reduce noise: By combining median filtering and first-order lag filtering, the present invention can make the sensor output signal clearer and more accurate. Median filtering is used to process normal signals, and first-order lag filtering is used to stabilize signals close to the threshold, effectively reducing noise and false alarms, and improving the robustness of the system. This is of great significance for some application scenarios with high measurement accuracy requirements, such as industrial, medical, and environmental protection fields.

[0052] Third, small size of the hardware circuit board: The median filter, first-order lag filter, Gaussian filter, and Savitzky-Golay smoothing filter in the present invention are all functional modules in the program sense, rather than product components in the physical sense. And the method corresponding to the present invention can reduce the power consumption of the sensor and improve the utilization efficiency of the sensor, which enables the present invention to adopt a miniaturized design for the sensor array and its corresponding hardware circuit board. Specifically, the size of the hardware circuit board with a miniaturized design is similar to that of a one-yuan coin with a diameter of 25 mm, and is particularly suitable for devices with limited space and portable electronic products.

[0053] Fourth, reduce system cost: The present invention selects friendly algorithms such as the KNN algorithm and the baseline calibration algorithm. While realizing complex functions, it maintains hardware friendliness, reduces the difficulty and cost of hardware implementation, is convenient for practical application and promotion, and helps with large-scale mass production and commercialization.

[0054] Fifth, enhance long-term stability: For the MEMS gas sensor array, by adopting the adaptive mean baseline calibration algorithm, the present invention successfully solves the drift problem of the gas sensor, and improves the long-term stability and measurement accuracy of the sensor.

[0055] Sixth, strong universality. The technical effects in the above aspects also make the present invention have the advantage of strong universality. Description of the Drawings

[0056] Figure 1 is the schematic diagram of the flexible sensor array circuit in the present invention;

[0057] Figure 2 is the hardware circuit framework diagram of the present invention;

[0058] Figure 3 is the schematic diagram of the gas sensor array and its circuit in the present invention;

[0059] Figure 4 is the flow chart of the gesture recognition algorithm in the present invention;

[0060] Figure 5 is the flow chart of the baseline calibration algorithm in the present invention. Specific embodiments

[0061] In order to facilitate those skilled in the art to fully understand the technical solution of the present invention, the following introduces two embodiments of the present invention in conjunction with the accompanying drawings.

[0062] Embodiment 1

[0063] As Figure 1 , Figure 2 shown, a flexible sensor array information detection system (hereinafter referred to as the system for short) includes a hardware circuit board with a microprocessor as the core, as well as a flexible sensor array, a power supply, a power management unit, and a core microprocessor circuit installed on the hardware circuit board; the power supply is electrically connected to the core microprocessor circuit and the sensor array through the power management unit; the sensor array information detection system further includes an embedded application program installed in the microprocessor, an application program corresponding to the embedded application program installed in a mobile phone or a computer, and a computer software program for controlling the operation of the entire sensor array information detection system. The embedded application program is provided with a median filtering function module (median filter), a first-order lag filtering function module (first-order lag filter), a Savitzky-Golay smoothing filtering function module (smoothing filter), and a Gaussian filtering function module (Gaussian filter); the user's mobile phone or computer is communicatively connected to the microprocessor.

[0064] In this embodiment, the power supply is a DC power supply such as a lithium battery. The function of the power management unit is to provide a bias voltage for each flexible sensor (hereinafter referred to as the sensor) that makes up the flexible sensor array, thereby expanding the response range of the sensor and reducing the power consumption of the sensor. When the change amount of the sampling signal is less than a certain value, under the control of the embedded application program, the system enters the low-power mode, reducing the operating frequency of the sensor, thereby reducing the sampling rate of the sensor.

[0065] It should be noted that in this embodiment, a trigger threshold is set for the hardware circuit. Under the control of the computer software program, the system compares the sampled values of the signal with the optimal value within a period of time. If the sampled value is normal, median filtering is performed using a median filter; if the sampled value is always close to the trigger threshold, first-order lag filtering is performed using a first-order lag filter. Through the cooperation of median filtering and first-order lag filtering, the system can effectively suppress noise interference. In addition, under the control of the computer software program, the system encapsulates the collected data into data packets and inserts judgment bits and check bits to effectively prevent data loss.

[0066] It should be emphasized that the median filter, first-order lag filter, Gaussian filter, and Savitzky-Golay smoothing filter in this embodiment are all functional modules in the program sense, rather than product components in the physical sense. Moreover, the method corresponding to this embodiment can reduce the power consumption of the sensor and improve the utilization rate of the sensor, which enables the sensor array and the hardware circuit board in this embodiment to adopt a miniaturized design. Specifically, the design size of the hardware circuit board in this embodiment is similar to the size of a one-yuan coin with a diameter of 25 mm, thus greatly reducing the occupation of physical space and being particularly suitable for devices with limited space and portable electronic products.

[0067] Embodiment 2

[0068] The basic structure of Embodiment 2 is the same as that of Embodiment 1. The main difference between the two is that the sensor array in Embodiment 1 is a flexible sensor array, while the sensor array in Embodiment 2 is a MEMS gas sensor array. In addition, there are slight differences in the circuit structure between the two. For example, a decoupling capacitor is provided in the circuit of Embodiment 2 (see Figure 3 , Figure 3 adopted a horizontal layout, and the decoupling capacitor is located at Figure 3 the rightmost lower end in

[0069] ), and its function is to further improve the anti-noise interference ability during the operation of the system, while the circuit of Embodiment 1 is not configured with a decoupling capacitor.

[0070] Hereinafter, the design method of the present invention will be further introduced. The design method includes the following steps:

[0071] Step 1, design the overall architecture: As shown in Figure 1 , Figure 2 or Figure 3 ,Figure 2 As shown, it includes a sensor array, a core microprocessor circuit, an embedded application installed in the microprocessor, an application installed in a mobile phone or computer corresponding to the embedded application, and the design of a computer software program for controlling the operation of the entire sensor array information detection system;

[0072] Step 2, design the core circuit board: Select a hardware circuit board with a microprocessor as the core, and transmit and receive the sampled data via Bluetooth through digital filtering and data encapsulation.

[0073] Step 3, design the algorithms: Select the KNN algorithm and the baseline calibration algorithm. The KNN algorithm (fully named K-Nearest Neighbors algorithm, also known as the K-nearest neighbor algorithm) is an instance-based learning algorithm, and its core idea is to perform classification or regression prediction by measuring the distances between different data points. The process of the baseline calibration algorithm is as Figure 5 shown, and its core idea is to convolve the data using a Gaussian function to achieve the purpose of smoothing the data. In the present invention, the standard deviation of the Gaussian filter is set to 100, which means that the filter will give a larger weight to the current data point and the data within its neighborhood, while the weight will gradually decrease for the data far from the current point.

[0074] Step 4, design the signal analysis: A method of adding a threshold determination mechanism to the KNN algorithm is proposed, that is, the T-KNN method, which compares the data with a predetermined threshold and eliminates the signals below the threshold, thereby effectively reducing the consumption of hardware resources.

[0075] Step 5, optimize the performance: The present invention proposes an adaptive mean baseline calibration strategy, selects a suitable time window and calculates the average value of the data within it. The length of this window is selected based on the change characteristics of the signal and the sampling frequency to ensure that it accurately reflects the baseline trend of the signal. Next, the original data is adjusted by subtracting the average value of the window to track the baseline change and obtain the corrected signal.

[0076] It should be noted that when implementing the algorithms designed in Step 3, the following several issues should also be noted and corresponding technical means should be adopted:

[0077] First, different features or variables often have different dimensions and orders of magnitude, which may cause the effects of certain features to be over-amplified or reduced during comprehensive analysis, thus affecting the accuracy of the results. To eliminate the influence of dimensions on data analysis and ensure that each feature has the same weight and importance in subsequent processing, the present invention adopts range normalization (that is, maximum-minimum normalization, hereinafter simply referred to as normalization) to transform the data into a specific range (between 0 and 1) to eliminate the dimensional differences between the data. The normalization formula is as follows:

[0078]

[0079] wherein, normalized data represents the normalized data, and raw data represents the raw data, min value represents the minimum value in the data, and max value represents the maximum value in the data.

[0080] Second, in order to reduce the random noise in the data, especially the noise caused by burst interference or errors, the present invention adopts a median filter, and its core principle is the sorting statistic theory. During the processing, the median filter examines all the values within a specific window in the data sequence and sorts these values by size. Then, the filter selects the value located in the middle position after sorting as the output, that is, the median of all the values within the window. The window size is set to 501, which means that each polynomial fitting is based on 501 data points before and after the current data point (including the current point). Such a setting helps to effectively remove random noise and outliers while retaining the main features of the data.

[0081] Third, the present invention divides the entire data sequence into several periods. These periods can be based on time, frequency or other related characteristics. The data within each period is considered to be relatively stable and has a similar baseline offset. The present invention divides the data into several periods and calculates the average value within each period as the baseline baseline. Then, the corresponding baseline value is subtracted from the raw data to obtain the corrected data as follows:

[0082] corrected data = normalized data - baseline

[0083] Fourth, in order to further smooth the data and retain its key signal features, the present invention selects a Savitzky-Golay smoothing filter. This filter is particularly suitable for the field of signal processing, where it is necessary to keep the shape and width of the signal unchanged while smoothing the noise. The window length and polynomial order are selected to balance noise suppression and signal feature retention. The window length is set to 301, which means that each polynomial fitting is based on 301 data points before and after the current data point (including the current point). The polynomial order is selected as 3, that is, a cubic polynomial is used for fitting. Such a setting helps to better retain the main features and peak shape of the signal while smoothing the data.

[0084] Fifthly, in order to further improve the smoothness of the data, the present invention adopts a Gaussian filter. Its core idea is to convolve the data with a Gaussian function to achieve the purpose of smoothing the data. In this algorithm, the present invention sets the standard deviation of the Gaussian filter to 100, which means that the filter will give a larger weight to the current data point and the data within its neighborhood, while for the data far from the current point, the weight will gradually decrease.

[0085] The above describes the structural features and design methods of the present invention. Next, the gesture recognition algorithm process in the present invention will be further introduced. As Figure 4 shown, the gesture recognition algorithm process includes the following steps:

[0086] S1. Discriminate the original data set: After obtaining the original gesture action data from the lower computer, first perform data discrimination. If the small threshold ≤ data ≤ large threshold, the data is determined to be valid; if the data < small threshold or the data > large threshold, the data is determined to be invalid.

[0087] S2. Initialize the gesture sample data set: Mark the data after discrimination, and divide the total samples into a training sample set and a test sample set, and label them according to the data of different people in the collected samples and different actions.

[0088] S3. Calculate the Euclidean distance: That is, the distance of the straight line between two points. By calculating the real distance between two points in the m-dimensional space or the natural length of the vector (i.e., the distance from this point to the origin), the nearest neighbor of a given sample can be identified. The specific calculation formula is as follows:

[0089]

[0090] In the formula, d is the distance between two points U and V, and x iu and x iv are m features on the sample data.

[0091] S4. Sort from small to large: According to the calculated Euclidean distances, sort them from small to large, and each data can be mapped to a label.

[0092] S5. Take the labels of the first K data: According to the nearest neighbor idea, circle the K objects with the closest distance as the nearest neighbors of the test object. Determine the final sample category through a majority vote.

[0093] S6. Obtain the result of gesture recognition: Match and connect with the interface of the upper computer software, and output the result of the majority vote.

[0094] It should be further noted that the above gesture recognition algorithm is designed by the present invention for deaf-mute people. When used in a system equipped with a strain sensor array, the gesture accuracy rate reaches 99.8%.

[0095] In summary, by implementing a variety of low-power and noise reduction strategies, such as multi-bias voltages, self-awakening mechanisms, average hysteresis filtering, and data encapsulation, as well as hardware-friendly algorithms tailored for specific sensor arrays, such as the KNN algorithm combined with threshold discrimination and the adaptive mean baseline calibration algorithm, the present invention significantly improves the performance, stability, and accuracy of the sensors, achieving highly accurate recognition of the gestures of deaf-mute people and effective compensation for the drift of the MEMS gas sensor array.

Claims

1. A sensor array information detection system includes a hardware circuit board with a microprocessor as the core, as well as a sensor array, a power supply, a power management unit, and a core microprocessor circuit installed on the hardware circuit board; the power supply is electrically connected to the core microprocessor circuit and the sensor array through the power management unit; it is characterized in that: The power management unit provides bias voltages for the sensors that make up the sensor array; the sensor array information detection system further includes an embedded application installed in the microprocessor, an application corresponding to the embedded application installed in a mobile phone or a computer, and a computer software program that controls the operation of the entire sensor array information detection system. The embedded application is provided with a median filtering function module, a first-order lag filtering function module, a Savitzky-Golay smoothing filtering function module, and a Gaussian filtering function module; the user's mobile phone or computer is communicatively connected to the microprocessor.

2. The sensor array information detection system according to claim 1, wherein: The sensor array is a flexible sensor array.

3. The sensor array information detection system according to claim 1, wherein: The sensor array is a strain sensor array.

4. The sensor array information detection system according to claim 1, wherein: The sensor array is a gas sensor array.

5. The sensor array information detection system according to claim 4, wherein: The gas sensor array is a MEMS gas sensor array.

6. A design method for a sensor array information detection system according to any one of claims 1-5, characterized in that, It includes the following steps: Step 1, design the overall architecture: including the design of the sensor array, the core microprocessor circuit, the embedded application installed in the microprocessor, the application corresponding to the embedded application installed in a mobile phone or a computer, and the computer software program that controls the operation of the entire sensor array information detection system; Step 2, design the core circuit board: select a hardware circuit board with a microprocessor as the core, and transmit and receive the sampled data through Bluetooth by digital filtering and data encapsulation; Step 3, design the algorithm: select the KNN algorithm and the baseline calibration algorithm; Step 4, design the signal analysis: add a threshold determination mechanism to the KNN algorithm, that is, the T-KNN method, compare the data with a predetermined threshold, and eliminate the signals below the threshold; Step 5, optimize the performance: select an appropriate time window and calculate the average value of the data within it.

7. The design method of the sensor array information detection system according to claim 6, characterized in that: When the sensor array in Step 1 is a MEMS gas sensor array, the baseline calibration algorithm in Step 3 is an adaptive mean baseline calibration algorithm.

8. The design method of the sensor array information detection system according to claim 7, characterized in that: A gesture recognition algorithm process is set in the computer software program in Step 1. The gesture recognition algorithm process includes the following steps: S1, discriminate the original data set: after obtaining the original gesture action data from the lower computer, first discriminate the data. If the small threshold ≤ data ≤ the large threshold, the data is determined to be valid. If the data < the small threshold or the data > the large threshold, the data is determined to be invalid; S2, initialize the gesture sample data set: mark the data after discrimination, and divide the total samples into a training sample set and a test sample set, and label them according to the data of different people in the collected samples and different actions; S3, calculate the Euclidean distance: that is, the straight-line distance between two points. By calculating the true distance between two points in an m-dimensional space, or the natural length of a vector (i.e., the distance from this point to the origin), the nearest neighbor of a given sample can be identified. The specific calculation formula is as follows: where d is the distance between two points U and V, and x iu and x iv are m features on the sample data; S4, sort from small to large: according to the calculated Euclidean distances, sort them from small to large, and each data can be mapped to a label; S5, Obtain the labels of the top K data: According to the nearest neighbor idea, circle the K objects with the closest distance as the neighbors of the test object. Determine the final sample category through majority voting; S6, Obtain the result of gesture recognition: Match and connect with the interface of the host computer software, and output the result of majority voting.

Images