Smell component analysis method and device based on multi-sensor array

Abstract

The invention discloses an odor component analysis method and device based on a multi-sensor array. The method comprises the following steps: 1, acquiring an initial signal set through the multi-sensor array exposed to a gas sample; 2, performing signal conditioning on the initial signal set to generate a conditioned signal set; performing dynamic error correction on the conditioned signal set based on temperature and humidity data of an environment where the odor sample is located, and generating an environment compensated signal set; 3, performing pattern recognition on the signal set after environment compensation to obtain odor component-concentration relation data of the gas sample; the device comprises a multi-sensor array, a data preprocessing unit and a power supply module, wherein the data preprocessing unit comprises a microprocessor, a signal conditioning circuit and a temperature and humidity sensor. Therefore, the design can accurately analyze the smell components and the corresponding concentration information.

Description

TECHNICAL FIELD

[0001] The present application relates to an odor component analysis method and device, belonging to the technical field of odor data processing, in particular to an odor component analysis method and device based on a multi-sensor array. BACKGROUND

[0002] In the current research and application of odor components, accurate component collection and analysis are the key to realizing odor digitization and standardization. However, the existing technology has obvious limitations: first, the commonly used single type sensor can only respond to specific categories of odor molecules, making it difficult to fully capture multiple components in complex odors, resulting in incomplete information collection and affecting the accuracy of analysis; second, the sensor signal is easily disturbed by changes in environmental temperature and humidity, resulting in a common error of more than 15% in the odor components and concentration data analyzed, also affecting the accuracy of analysis; the above problems together result in the inability to accurately analyze odor components and corresponding concentration information.

[0003] The Chinese patent application with application number 202110733541.5 and application date June 30, 2021 discloses a multi-gas detection method and system for a MEMS gas sensor array, which method includes: constructing a gas sensor network of a preset MEMS gas sensor array, receiving detection data of each first gas sensor in the preset MEMS gas sensor array in a target environment through the gas sensor network, determining the first gas components and the respective gas concentrations in the target environment according to the detection data, collecting the gas in the target environment and storing it in a preset container, using a second gas sensor in the preset MEMS gas sensor array to perform a light absorption test on the stored gas in the preset container, obtaining a test result, determining the second gas components and the respective gas concentrations in the target environment according to the test result, and comprehensively obtaining the target gas components and the respective gas concentrations in the target environment according to the first and second gas components and the respective gas concentrations. Although this patent uses a MEMS gas sensor array to improve the accuracy of the final detection result, it still has the following defects:

[0004] This design first detects with a sensor array and then samples for optical testing, which cannot truly reflect the complete composition of the odor, still resulting in the inability to accurately analyze odor components and corresponding concentration information.

[0005] The information disclosed in this BACKGROUND section is only intended to increase an understanding of the general context of the present patent application and should not be taken as an acknowledgement or any form of suggestion that this information constitutes prior art that is publicly known. SUMMARY

[0006] The purpose of the present application is to overcome the defects and problems in the prior art that cannot accurately analyze the odor component and the corresponding concentration information, and to provide an odor component analysis method and device based on a multi-sensor array which can accurately analyze the odor component and the corresponding concentration information.

[0007] To achieve the above purpose, the technical solution of the present application is: an odor component analysis method and device based on a multi-sensor array, the method comprising:

[0008] First step: obtaining an initial signal set through a multi-sensor array exposed to an odor sample;

[0009] Second step: first, signal conditioning is performed on the above initial signal set to generate a conditioned signal set; then, based on the temperature and humidity data of the environment where the odor sample is located, dynamic error correction is performed on the conditioned signal set to generate an environment-compensated signal set;

[0010] Third step: pattern recognition is performed on the above environment-compensated signal set to obtain the odor component-concentration relationship data of the odor sample.

[0011] The method further comprises:

[0012] Fourth step: converting the above odor component-concentration relationship data into a standard format to generate a standardized data package.

[0013] In the first step, the multi-sensor array comprises a plurality of electrochemical sensors, a plurality of optical sensors, and a plurality of semiconductor sensors.

[0014] In the second step, the signal conditioning refers to amplification and denoising processing in sequence.

[0015] In the second step, the dynamic error correction of the conditioned signal set based on the temperature and humidity data of the environment where the odor sample is located to generate an environment-compensated signal set refers to: first, calling a pre-stored temperature and humidity-signal error mapping relationship, calculating the corresponding signal compensation amount according to the real-time collected temperature and humidity data of the environment where the odor sample is located; then, using the signal compensation amount to correct the conditioned signal set to generate an environment-compensated signal set.

[0016] The pre-stored temperature and humidity-signal error mapping relationship is obtained by the following method:

[0017] Before obtaining the initial signal set, the response signal of the multi-sensor array to standard gas is measured in environments with a plurality of different temperature and humidity combinations, and then based on the deviation of the response signal and the reference signal under the different temperature and humidity, the temperature and humidity-signal error mapping relationship is established through data fitting.

[0018] The third step is to perform pattern recognition on the above-mentioned environment-compensated signal set to obtain the odor component-concentration relationship data of the odor sample, which refers to: first, performing feature extraction on the above-mentioned environment-compensated signal set to obtain a feature vector; then, identifying the category identifier of the odor component in the odor sample through a classification model by using the feature vector; then, calculating the concentration value of the corresponding odor component based on the category identifier and the feature vector through a category odor component regression model; and finally, combining the concentration value of each odor component with the corresponding category identifier to generate the odor component-concentration relationship data of the odor sample.

[0019] The classification model is obtained by the following method:

[0020] First, a plurality of standard odor samples with known category identifiers are collected, and standard feature vectors corresponding to the standard odor samples are extracted. Then, a machine learning algorithm is supervised trained by taking the standard feature vectors as input and the corresponding category identifiers as expected output, so as to form a mapping relationship from the feature vector to the category identifier, and finally obtain the classification model.

[0021] The category odor component regression model is obtained by the following method:

[0022] A dedicated regression model is established for each category of odor component. For each category of odor component, a plurality of standard samples with known concentration values are collected, and corresponding standard feature vectors are extracted. Then, a regression algorithm is supervised trained by taking the feature vectors as input and the corresponding concentration values as expected output, so as to form a mapping relationship from the feature vector to the concentration value, and finally obtain the dedicated regression model of the category odor component.

[0023] The device comprises a multi-sensor array, a data preprocessing unit, and a power supply module.

[0024] The data preprocessing unit comprises a microprocessor, a signal conditioning circuit, and a temperature and humidity sensor.

[0025] The input end of the signal conditioning circuit is connected with the multi-sensor array, the output end of the signal conditioning circuit is connected with the microprocessor, and the output end of the temperature and humidity sensor is connected with the microprocessor.

[0026] The power supply module is connected with the multi-sensor array, the microprocessor, the signal conditioning circuit, and the temperature and humidity sensor, respectively.

[0027] The device further comprises a housing.

[0028] The multi-sensor array, the data preprocessing unit, and the power supply module are all arranged in the housing. The front end of the housing is provided with an opening, the sensing surface of the multi-sensor array is arranged at the opening, and the opening is provided with a protective cover.

[0029] Compared with the prior art, the present application has the following advantages:

[0030] 1. The present application is a kind of smell component analysis method and device based on multi-sensor array, the method comprises the following steps: obtaining the initial signal set of the smell sample through the multi-sensor array first, then sequentially performing signal conditioning and dynamic error correction on the initial signal set to generate the environment compensated signal set, and then performing pattern recognition on the environment compensated signal set to obtain the smell component-concentration relationship data of the smell sample. The present application has the following advantages:

[0031] Firstly, the initial signal set of the smell sample is obtained through the multi-sensor array, and then the signal conditioning and dynamic error correction are sequentially performed to generate the environment compensated signal set, and then the pattern recognition is performed to obtain the smell component-concentration relationship data of the smell sample.

[0032] Secondly, the chemical information of different smell molecules in the smell sample is converted into the initial signal set that can be analyzed through the cross-response characteristics of the multi-sensor array, which ensures the synchronous and comprehensive acquisition of the information of multiple components in the smell sample, and overcomes the defect of incomplete acquisition caused by the limited response range of a single sensor.

[0033] Thirdly, the conditioned signal set is corrected through real-time monitoring and dynamic error compensation of the environment temperature and humidity, which effectively eliminates the influence of environmental fluctuations on the analysis, makes the identification of the smell component more reliable, and the calculation of the concentration value more accurate, so that the analysis error is less than or equal to 5%, and the analysis error is significantly reduced.

[0034] Therefore, the present application can not only analyze the smell data, but also accurately analyze the smell component and the corresponding concentration information.

[0035] 2. In the present application, the signal conditioning refers to sequentially performing amplification and denoising processing. When applied, the amplitude and signal-to-noise ratio of the initial signal set are effectively improved through signal amplification processing, so that the subsequent processing link can effectively identify and quantify the subtle concentration change of the smell. At the same time, the inherent noise of the circuit and the environmental electromagnetic interference are suppressed through denoising processing, which ensures the purity and authenticity of the signal characteristics, and provides a high-quality signal basis for the subsequent pattern recognition algorithm to accurately analyze the smell component and its concentration. Therefore, the present application can not only accurately analyze the smell component and the corresponding concentration information, but also improve the quality and reliability of the initial signal set.

[0036] 3、The method and device for analyzing odor components based on a multi-sensor array, the method further comprises: in the fourth step, converting the odor component-concentration relationship data into a standard format to generate a standardized data package; in application, the odor component-concentration relationship data is packaged into a structured data package including standard chemical codes, unified measurement units and complete metadata, so as to ensure the universality and consistency of the data format, enable the standardized data package to be recognized and analyzed by different terminal devices and analysis systems, and thus realize efficient circulation and sharing of odor data in a cross-platform and cross-system environment. Therefore, the quality and reliability of the initial signal set are improved, and the data circulation efficiency is improved.

[0037] 4、The method and device for analyzing odor components based on a multi-sensor array, the device further comprises a housing, the multi-sensor array, the data preprocessing unit and the power supply module are arranged in the housing; an opening is arranged at the front end of the housing, the sensing surface of the multi-sensor array is arranged at the opening, and the opening is provided with a protective cover; in application, the integrated housing design and the protective cover structure provide a multiple protection mechanism for the device, first, the multi-sensor array, the data preprocessing unit and the power supply module are integrated in the housing, so as to effectively resist physical damage and pollution of the internal elements by the external environment, at the same time, the protective cover at the front end provides targeted protection for the multi-sensor array in a non-working state, so as to avoid pollution or mechanical damage of sensitive elements, and when working, opening the protective cover can ensure that the sensing surface of the sensor is in full contact with the environment to be measured, so as to ensure effective collection of odor molecules, thereby improving the durability and reliability of the device. Therefore, the data circulation efficiency is improved, and the durability and reliability of the device are improved. BRIEF DESCRIPTION OF DRAWINGS

[0038] Figure 1 is a flowchart of the present application.

[0039] Figure 2 is a schematic diagram of the overall structure of the present application.

[0040] Figure 3 is a schematic diagram of the internal module connection of the device in the present application.

[0041] In the figure: multi-sensor array 1, data preprocessing unit 2, microprocessor 21, signal conditioning circuit 22, temperature and humidity sensor 23, power supply module 3, power supply 31, data output module 4, wired interface 41, housing 5, opening 51, protective cover 52, charging interface 53, and power indicator 54. DETAILED DESCRIPTION

[0042] The present application is further described in detail below in conjunction with the accompanying drawings and specific embodiments.

[0043] Reference Figure 1- Figure 3 A method and device for analyzing odor components based on a multi-sensor array, the method comprising:

[0044] Step 1: obtaining an initial signal set by exposing a multi-sensor array to an odor sample;

[0045] Step 2: first, performing signal conditioning on the initial signal set to generate a conditioned signal set; then, based on the temperature and humidity data of the environment in which the odor sample is located, performing dynamic error correction on the conditioned signal set to generate an environment-compensated signal set;

[0046] Step 3: performing pattern recognition on the environment-compensated signal set to obtain odor component-concentration relationship data of the odor sample.

[0047] The method further comprises:

[0048] Step 4: converting the odor component-concentration relationship data into a standard format to generate a standardized data package.

[0049] In the first step, the multi-sensor array comprises a plurality of electrochemical sensors, a plurality of optical sensors, and a plurality of semiconductor sensors.

[0050] In the second step, the signal conditioning refers to sequentially performing amplification and denoising.

[0051] In the second step, the dynamic error correction of the conditioned signal set based on the temperature and humidity data of the environment in which the odor sample is located to generate an environment-compensated signal set refers to: first, calling a pre-stored temperature and humidity-signal error mapping relationship, calculating the corresponding signal compensation amount according to the real-time collected temperature and humidity data of the environment in which the odor sample is located; then, using the signal compensation amount to correct the conditioned signal set to generate an environment-compensated signal set.

[0052] The pre-stored temperature and humidity-signal error mapping relationship is obtained by the following method:

[0053] Before obtaining the initial signal set, the response signals of the multi-sensor array to standard gases are measured in environments with a plurality of different temperature and humidity combinations, and then based on the deviations of the response signals and reference signals under different temperatures and humidities, the temperature and humidity-signal error mapping relationship is established through data fitting.

[0054] In the third step, the mode recognition of the environment-compensated signal set is performed to obtain the odor component-concentration relationship data of the odor sample, which refers to: first, feature extraction is performed on the environment-compensated signal set to obtain a feature vector; then, the feature vector is input into a classification model to identify the category label of the odor component in the odor sample; then, based on the category label and the feature vector, a category odor component regression model is used to calculate the concentration value of the corresponding odor component; finally, the concentration value of each odor component is associated with the corresponding category label to generate the odor component-concentration relationship data of the odor sample.

[0055] The classification model is obtained by the following method:

[0056] First, a plurality of standard odor samples with known category labels are collected, and standard feature vectors corresponding to the standard odor samples are extracted. Then, the standard feature vectors are input into a machine learning algorithm, and the corresponding category labels are expected as the output. The machine learning algorithm is supervised trained to form a mapping relationship from the feature vector to the category label, and finally the classification model is obtained.

[0057] The category odor component regression model is obtained by the following method:

[0058] A dedicated regression model is established for each category of odor component. For each category of odor component, a plurality of standard samples with known concentration values are collected, and the corresponding standard feature vectors are extracted. Then, the feature vectors are input into a regression algorithm, and the corresponding concentration values are expected as the output. The regression algorithm is supervised trained to form a mapping relationship from the feature vector to the concentration value, and finally the dedicated regression model of the category odor component is obtained.

[0059] The device comprises a multi-sensor array 1, a data preprocessing unit 2, and a power supply module 3.

[0060] The data preprocessing unit 2 comprises a microprocessor 21, a signal conditioning circuit 22, and a temperature and humidity sensor 23.

[0061] The input end of the signal conditioning circuit 22 is connected to the multi-sensor array 1, the output end of the signal conditioning circuit 22 is connected to the microprocessor 21, and the output end of the temperature and humidity sensor 23 is connected to the microprocessor 21.

[0062] The power supply module 3 is connected to the multi-sensor array 1, the microprocessor 21, the signal conditioning circuit 22, and the temperature and humidity sensor 23, respectively.

[0063] The device further comprises a housing 5.

[0064] The multi-sensor array 1, the data preprocessing unit 2 and the power supply module 3 are arranged in the shell 5; the shell 5 is provided with an opening 51 at the front end, and the sensing surface of the multi-sensor array 1 is arranged at the opening 51; and the opening 51 is provided with a protective cover 52.

[0065] The application is further described as follows:

[0066] The multi-sensor array 1 is arranged at the front end of the shell 5 through a fixing support, and the sensing surface of each sensor in the multi-sensor array 1 is exposed outside, so that the multi-sensor array 1 is convenient to contact the odor source.

[0067] In the first step, the multi-sensor array includes a plurality of sensors of different types, and the number of each type of sensor is multiple.

[0068] The power supply module 3 includes a power supply 31, which is a rechargeable lithium battery with a capacity of 2000-5000mAh, and is used to provide power support for the multi-sensor array 1, the microprocessor 21, the signal conditioning circuit 22, the temperature and humidity sensor 23 and the data output module 4.

[0069] The device further includes a data output module 4, which includes a wireless communication module and a wired interface 41; and is used to transmit the standardized data packet to an external intelligent terminal (such as a mobile phone or a computer) or a storage device.

[0070] Preferably, the wireless communication module in the data output module 4 is Bluetooth or Wi-Fi.

[0071] Preferably, the wired interface 41 is a USB interface.

[0072] In the first step, the multi-sensor array exposed to the odor sample is used to obtain an initial signal set, which means that the multi-sensor array is arranged at a position within a distance of 5-10cm from the target odor source to collect the odor sample; each type of sensor in the multi-sensor array works in parallel based on different physical and chemical principles, produces specific responses to different types of molecules in the odor sample, generates electrical signals corresponding to the characteristics of each odor component, and collectively forms the initial signal set.

[0073] Preferably, the multi-sensor array is arranged at a position within a distance of 5-10cm from the target odor source, because this distance range can ensure that the sensor array is in a diffusion layer with appropriate odor molecule concentration, which can avoid the problems of sensor saturation or pollution caused by too close distance, and prevent the odor molecules from being excessively diluted and cross-contaminated caused by too far distance, so as to ensure that the collected signal has high signal-to-noise ratio, stability and authenticity.

[0074] In the second step of the application, the temperature and humidity data of the environment where the odor sample is located are collected in real time by the temperature and humidity sensor 23 to obtain the temperature and humidity values of the environment where the odor sample is located.

[0075] Embodiment 1

[0076] Referring to Figure 1 — Figure 3 A method and device for analyzing odor components based on a multi-sensor array, the method comprising:

[0077] First step: obtaining an initial signal set by exposing a multi-sensor array to an odor sample;

[0078] Second step: first, signal conditioning is performed on the initial signal set to generate a conditioned signal set; then, based on the temperature and humidity data of the environment where the odor sample is located, dynamic error correction is performed on the conditioned signal set to generate an environment-compensated signal set;

[0079] Third step: performing pattern recognition on the environment-compensated signal set to obtain odor component-concentration relationship data of the odor sample.

[0080] Embodiment 2

[0081] The basic content is the same as that of Embodiment 1, except that the method further comprises:

[0082] Fourth step: converting the odor component-concentration relationship data into a standard format to generate a standardized data package.

[0083] In application, the odor component-concentration relationship data generated in the third step is encapsulated according to the structure of the standard format to generate a standardized data package; specifically, the standard format is "data header | environmental parameters | component data | check code"; first, the data header is constructed, and the version number and device unique identifier that meet the standard are filled in; then, the temperature and humidity data recorded during collection are entered as the environmental parameter segment; next, the name of each odor component is converted into a globally unique digital code by querying a standard odor component dictionary, and is paired with the concentration value to form the component data segment in the order of the code; finally, the check code is calculated for the entire data package content and appended to the end to complete the generation of the standardized data package; the data package format is uniform, which is conducive to standardized storage and cross-platform transmission, and establishes a unified and reliable data foundation for subsequent analysis, sharing and application of data.

[0084] Embodiment 3

[0085] The basic content is the same as that of Embodiment 1, except that in the first step, the multi-sensor array comprises a plurality of electrochemical sensors, a plurality of optical sensors, and a plurality of semiconductor sensors.

[0086] In application, multiple sensors of each type are provided to improve the robustness and reliability of the system through the dual mechanisms of data redundancy and cross-validation. On the one hand, multiple sensors of the same type work in parallel to provide necessary data redundancy, which can effectively eliminate abnormal readings caused by instantaneous sensor failure or environmental noise by using median selection or mean calculation algorithms. On the other hand, multiple sensors of the same type form a cross-validation mechanism. When most sensors produce consistent responses to a specific odor molecule, and individual sensors produce abnormal responses, they can be determined as interference or failure signals according to the majority principle, thereby reducing the false positive rate and false negative rate of the system from the algorithm level, and ensuring the accuracy of the detection results.

[0087] Preferably, the number of sensors of each type in the multi-sensor array is set to 3-7.

[0088] In the specific detection process, various sensors achieve specific responses to odor molecules in the odor sample based on different physical and chemical principles. The electrochemical sensor works based on the oxygen reduction reaction principle. When electrically active molecules such as alcohols and aldehydes diffuse to the surface of the working electrode of the electrochemical sensor, redox reaction occurs, producing a proportional electron transfer to the gas concentration, outputting a microampere-level current signal. The optical sensor works based on the photoionization detection principle. By emitting ultraviolet light of a specific wavelength, volatile organic compounds such as esters and ketones are ionized. The ion current intensity is detected by the collection electrode, and the corresponding voltage signal is output. The semiconductor sensor works based on the surface adsorption effect. When terpene molecules are adsorbed on the surface of the metal oxide semiconductor, electron exchange occurs between the surface oxygen species, causing a significant change in the resistance value of the semiconductor sensor. The change is converted into a corresponding voltage signal output by the measurement circuit.

[0089] The three types of sensors work in parallel based on different detection principles, forming a complementary detection system. When exposed to an odor sample, each sensor starts collecting simultaneously and outputs response signals representing the characteristics of different types of molecules, forming an initial signal set containing multiple modal physical quantities. This signal set records the current signal sequence output by the electrochemical sensor, the voltage signal sequence output by the optical sensor, and the voltage signal sequence corresponding to the resistance change of the semiconductor sensor. Through the coordinated response of different types of molecules in complex odor mixtures and signal cross-validation, the sensitivity of the detection is improved to the 0.1-1 ppm level, and the cross-interference rate is controlled below 5%, providing a high-quality data basis for subsequent signal processing and component analysis. The multi-modal signals include current, voltage, and resistance change.

[0090] Example 4:

[0091] The basic content is the same as in embodiment 1, except that in the first step, the multi-sensor array is composed of at least three different types of sensors.

[0092] In application, the multi-sensor array at least includes a plurality of electrochemical sensors, a plurality of optical sensors, and a plurality of semiconductor sensors; in actual application, the sensor combination can also be expanded or adjusted according to the chemical molecular characteristics of the target detected odor and the needs of the specific application scene, for example, a quartz crystal microbalance (QCM) sensor is added to enhance the adsorption detection sensitivity of high molecular weight volatile organic compounds, or a conductive polymer sensor is used to improve the response speed of specific molecules such as ammonia and amine; through this configurable sensor array design, it is ensured that the device can flexibly and accurately cover the key component spectrum expected to exist in the target odor sample.

[0093] Embodiment 5:

[0094] The basic content is the same as in embodiment 1, except that in the second step, the signal conditioning refers to: sequentially performing amplification and denoising processing.

[0095] In application, first, the initial signal set is proportionally amplified by a precision amplifier in the signal conditioning circuit, so that its dynamic range is adapted to the input requirements of the subsequent analog-to-digital converter; this operation not only improves the signal-to-noise ratio of the system, but also ensures that the electrical signal representing the subtle changes in odor concentration has sufficient quantitative resolution; then, a filtering technology based on frequency domain characteristics is used for denoising; by configuring an electronic filter with specific cutoff characteristics (such as low-pass or band-pass type), the difference in frequency domain distribution between effective signals and noise is used to selectively suppress out-of-band noise components; this processing method can effectively purify the signal spectrum, retain the core frequency band representing the essential characteristics of the odor components, and thus ensure the consistency and comparability of the detection results under different environmental conditions;

[0096] The reason for sequentially performing amplification and denoising processing on the initial signal set is that the initial signal set output by the multi-sensor array is usually in the order of millivolts to microvolts; such weak signals are easily disturbed by system background noise during transmission; at the same time, the initial signal set actually collected is the superposition of effective odor component response and various interference signals, including circuit thermal noise, environmental electromagnetic interference, and power supply harmonics; if no corresponding signal conditioning processing is performed, these interferences will directly affect the accurate extraction of the signal and the reliability of subsequent analysis; through the above-mentioned orderly signal conditioning processing flow, the real odor response signal can be effectively extracted from the complex background noise, laying a foundation for subsequent environmental compensation and pattern recognition.

[0097] Embodiment 6:

[0098] The basic content is the same as in embodiment 1, except that in the second step, the dynamic error correction is performed on the conditioned signal set based on the temperature and humidity data of the environment where the odor sample is located to generate an environment-compensated signal set, which refers to: first, calling a pre-stored temperature and humidity-signal error mapping relationship, calculating the corresponding signal compensation amount according to the temperature value and humidity value data of the environment where the odor sample is located collected in real time; then, using the signal compensation amount to correct the conditioned signal set to generate an environment-compensated signal set.

[0099] In application, the dynamic error correction process is implemented as follows: the microprocessor 21 reads the environmental parameters collected by the temperature and humidity sensor 23 in real time to obtain the temperature value T (unit: ℃) and humidity value H (unit: %RH) of the current environment; then, a pre-stored temperature and humidity-signal error mapping relationship is called, which is stored in the form of a two-dimensional lookup table and records the correction coefficients of the signals of various types of sensors under different temperature and humidity combinations; according to the current (T, H) values, the corresponding signal compensation amount ΔS is calculated through a bilinear interpolation algorithm; the specific compensation formula is: S_compensated=S_original×(1+ΔS), where S_original is the original reading of each sensor in the conditioned signal set, and S_compensated is the compensated signal value; this dynamic compensation mechanism can effectively overcome the problems of sensor baseline drift and sensitivity change caused by changes in environmental temperature and humidity, reduce the measurement error caused by temperature and humidity interference from more than 15% before compensation to within 3%, and significantly improve the measurement accuracy and stability of the device under different environmental conditions.

[0100] Embodiment 7:

[0101] The basic content is the same as in embodiment 1, except that the pre-stored temperature and humidity-signal error mapping relationship is obtained by the following method: before obtaining the initial signal set, first measure the response signals of the multi-sensor array to standard gas in a variety of different temperature and humidity combinations, and then based on the deviation of the response signals and reference signals under different temperatures and humidities, the temperature and humidity-signal error mapping relationship is established through data fitting.

[0102] In application, firstly, in a controllable environment test chamber, a temperature gradient (specifically 10℃, 20℃, 30℃, 40℃) and a humidity gradient (specifically 30%RH, 50%RH, 70%RH, 90%RH) covering the expected working range of the device are systematically set, thereby forming 16 different temperature and humidity combination test conditions; then, under each specific temperature and humidity combination condition, the multi-sensor array is continuously exposed to a standard gas sample with known concentration (such as 100ppm isobutylene standard gas), and after the response of each sensor stabilizes, the output signal is recorded as the actual measurement signal under the environment; at the same time, the response signal of the same standard gas measured in a standard laboratory environment (25℃, 50%RH) is taken as the baseline reference signal; then, for each sensor and each temperature and humidity combination, the relative deviation value ΔS' of the actual measurement signal relative to the baseline reference signal is calculated; finally, taking the temperature value T and the humidity value H of all test points as two-dimensional input, and taking the calculated relative deviation ΔS' as output, a binary surface fitting is performed by using the least squares method, thereby establishing a mathematical model capable of continuously representing the relationship between temperature, humidity and signal deviation, i.e. the temperature and humidity-signal error mapping relationship, and the model parameters are solidified and stored in the microprocessor 21.

[0103] Embodiment 8:

[0104] The basic content is the same as that of embodiment 1, except that in the third step, the pattern recognition of the above environment-compensated signal set to obtain the odor component-concentration relationship data of the odor sample is that: first, feature extraction is performed on the above environment-compensated signal set to obtain a feature vector; then, the feature vector is input into a classification model to identify the category identification of the odor components in the odor sample; then, based on the category identification and the feature vector, a category odor component regression model is used to calculate the concentration value of the corresponding odor component; finally, the concentration values of the odor components are associated with the corresponding category identification to generate the odor component-concentration relationship data of the odor sample.

[0105] In application, firstly, multi-dimensional features representing sensor response characteristics are extracted from the environment-compensated signal set to construct a comprehensive feature vector including time domain and frequency domain parameters; then, the feature vector is input into a classification model to identify the specific odor component categories contained in the odor sample, such as “ethyl acetate”, “geraniol” and other specific odor component categories; then, for each identified odor component category, a corresponding category odor component regression model is called to calculate the specific concentration value of each component based on the concentration-related features in the feature vector; finally, all the identified odor components and their concentration values are associated to generate complete odor component-concentration relationship data; this method realizes accurate analysis of complex odor samples through step-by-step identification and quantitative analysis;

[0106] The feature vector includes peak response intensity, steady-state response value, response curve area, rise time, recovery time, maximum response rate, frequency domain principal component amplitude, and signal root mean square value. The rise time, recovery time, maximum response rate, and frequency domain principal component amplitude mainly represent the physical and chemical properties of the odor molecules and constitute the characteristic fingerprint of odor species identification. The peak response intensity, steady-state response value, response curve area, and signal root mean square value are positively correlated with the number of odor molecules and provide a quantitative basis for concentration calculation.

[0107] Embodiment 9:

[0108] The basic content is the same as that in Embodiment 1, except that the classification model is obtained by the following method: first, a plurality of standard odor samples of known categories are collected, and standard feature vectors corresponding to the standard odor samples are extracted, then the standard feature vectors are input, and the corresponding category identification is expected output, the machine learning algorithm is supervised training, the mapping relationship from the feature vector to the category identification is formed, and finally the classification model is obtained; the category odor component regression model is obtained by the following method: a dedicated regression model is established for each category of odor component; wherein, for each category of odor component, a plurality of standard samples of known concentration values are collected, and the corresponding standard feature vectors are extracted, then the feature vectors are input, and the corresponding concentration values are expected output, the regression algorithm is supervised training, the mapping relationship from the feature vector to the concentration value is formed, and finally the dedicated regression model of the category odor component is obtained.

[0109] In application, first, a training data set including a plurality of known category standard odor samples is constructed, covering all odor categories expected to be detected by the device, such as floral, fruity, alcoholic, aldehyde, etc.; for each category of odor sample, a standard feature vector containing peak response intensity, steady-state response value, response curve area, rise time, recovery time, maximum response rate, frequency domain principal component amplitude, and signal root mean square value is extracted; a support vector machine algorithm is used for supervised training with the standard feature vector as input and the corresponding category identification as output to establish a classification model, and the model parameters are optimized through cross-validation to make the classification accuracy rate reach more than 95%;

[0110] At the same time, a dedicated regression model is established for each category of odor component: standard samples of different concentration gradients are prepared, and the corresponding feature vectors are extracted; a random forest regression algorithm is used to train with the concentration-related features (peak response intensity, steady-state response value, response curve area, and signal root mean square value) in the feature vector as input and the known concentration value as output; each regression model is evaluated by root mean square error to ensure that the concentration prediction error is controlled within 3%; finally, the trained classification model and each regression model are integrated into the microprocessor 21 to form a complete pattern recognition system.

[0111] Embodiment 10:

[0112] The basic content is the same as that in Embodiment 1, except that the device comprises a multi-sensor array 1, a data preprocessing unit 2, and a power supply module 3; the data preprocessing unit 2 comprises a microprocessor 21, a signal conditioning circuit 22, and a temperature and humidity sensor 23; an input end of the signal conditioning circuit 22 is connected with the multi-sensor array 1, an output end of the signal conditioning circuit 22 is connected with the microprocessor 21, and an output end of the temperature and humidity sensor 23 is connected with the microprocessor 21; and the power supply module 3 is connected with the multi-sensor array 1, the microprocessor 21, the signal conditioning circuit 22, and the temperature and humidity sensor 23, respectively.

[0113] In application, first, the device is placed in a monitoring position at a distance of 5-10 cm from a target odor source, each sensor in the multi-sensor array 1 synchronously collects odor molecules, and outputs an initial signal set including current and voltage signals; at the same time, the temperature and humidity sensor 23 collects temperature and humidity data of an environment where the target odor source is located in real time; the signal conditioning circuit 22 sequentially performs amplification and frequency domain filtering processing on the initial signal set, and improves signal quality; the microprocessor 21 reads the signal set after conditioning and the temperature and humidity data, calls a temperature and humidity compensation algorithm built in the microprocessor 21 to dynamically correct the signal set after conditioning, and eliminates environmental interference; then, the microprocessor 21 extracts time domain and frequency domain features from the signal set after environmental compensation, constructs a multi-dimensional feature vector, identifies odor components through a pre-stored classification model, and calculates concentration values of each component by using a special regression model; finally, the microprocessor 21 associates the identified odor components with the concentration values, and encapsulates them into a standard format data packet; the whole process is continuously completed under the support of the power supply module 3, and realizes full-automatic processing from odor collection to component analysis.

[0114] Embodiment 11:

[0115] The basic content is the same as that in Embodiment 10, except that the device further comprises a data output module 4, the data output module 4 is connected with an output end of the microprocessor 21, the data output module 4 is connected with the power supply module 3, and the data output module 4 is arranged in the shell 5.

[0116] In application, the standard format data packet is output to a terminal device through the data output module 4; the power supply module 3 supplies power for the data output module 4; and the shell 5 provides protection for the data output module 4.

[0117] Embodiment 12:

[0118] The basic content is the same as that of Example 1, except that the device further comprises a shell 5; the multi-sensor array 1, the data preprocessing unit 2, and the power supply module 3 are all arranged in the shell 5; the front end of the shell 5 is provided with an opening 51, and the sensing surface of the multi-sensor array 1 is arranged at the opening 51; and the opening 51 is provided with a protective cover 52.

[0119] In application, the shell 5 is made of light-weight high-strength ABS material, and has an overall size of 10 cm x 5 cm x 3 cm and a total weight of ≤100 g, so as to realize the portable design of the device; the protective cover 52 at the front end of the shell 5 is arranged in a movable structure, and can provide physical protection for the multi-sensor array 1 in a non-working state, so as to avoid contamination or mechanical damage of the sensor; when working, the protective cover 52 is opened, so that the sensing surface of the sensor is exposed to the environment to be measured, and effective contact and collection of odor molecules are ensured.

[0120] Example 13

[0121] The basic content is the same as that of Example 10, except that the shell 5 further comprises a charging interface 53 and an electric quantity indicator 54 arranged on the side wall of the shell 5, and a wired interface 41.

[0122] In application, the charging interface 53 and the electric quantity indicator 54 are integrated on the shell 5 to form a complete power management system; the charging interface 53 is electrically connected with the power supply module 3 to realize the charging function, and the electric quantity indicator 54 is electrically connected with the power supply module 3 to display the remaining electric quantity in real time; at the same time, the wired interface 41 arranged on the side wall of the shell 5 serves as a physical transmission channel of the data output module 4, and works cooperatively with the wireless communication module to realize the transmission of the generated “odor component-concentration” structured data to the intelligent terminal or the storage device; in addition, the surface of the shell 5 is further provided with anti-slip lines to ensure the stability and safety of the handheld operation.

[0123] Example 14

[0124] The basic content is the same as that of embodiment 1, except that the difference lies in that the collection and analysis process of the rose smell in the outdoor environment are specifically explained; the device protection cover 52 is opened and the power supply is started, and the temperature and humidity sensor 23 collects the environmental parameters (such as temperature 25 DEG C, humidity 60%) in real time; the multi-sensor array 1 at the front end of the device is aligned with the rose flowers, and the detection distance is kept at 5-10 cm; the electrochemical sensor in the multi-sensor array 1 produces specific response to alcohol molecules (such as citronellol, geraniol), the optical sensor identifies ester molecules, the semiconductor sensor detects terpene molecules, and each sensor synchronously outputs a response signal; the signal conditioning circuit 22 amplifies and filters the original response signal, and transmits the signal after removing the environmental noise to the microprocessor 21; the microprocessor 21 dynamically corrects the signal error according to the environmental temperature and humidity data, and analyzes the component concentration data (such as citronellol 35%, geraniol 25%, phenethyl alcohol 20%, and other 20%) of the rose smell; finally, the structured "odor component-concentration" data are transmitted to the mobile phone terminal through the Bluetooth module of the data output module 4, and the detection results are displayed and stored in real time by the special APP.

[0125] Embodiment 15:

[0126] The basic content is the same as that of embodiment 1, except that the difference lies in that the rapid detection and quality monitoring process of the bread smell in the food production environment are specifically explained; the staff holds the device into the food workshop, and the multi-sensor array 1 is close to the sample to be detected at the bread baking station, and the detection distance is controlled at about 8 cm; after the device is started, the multi-sensor array 1 synchronously collects the aroma molecules released by the bread, and the concentration data of the key aroma components (such as ethyl acetate, butanedione, etc.) are analyzed by the microprocessor 21 after signal conditioning and temperature and humidity compensation; the "odor component-concentration" relationship data analyzed are transmitted to the workshop control computer through the USB interface, and real-time comparison and evaluation are carried out by the professional analysis software; if the concentration of some components deviates from the standard range, the system can timely issue a warning to guide the production process adjustment, so as to realize the rapid closed-loop control of the bread aroma quality.

[0127] The above only describes the preferred embodiments of the present application, and the protection scope of the present application is not limited to the above embodiments, but any equivalent modification or change made by the ordinary skilled person in the art according to the disclosed content of the present application shall be included in the protection scope recorded in the claims.

Claims

1. A method for resolving odor components based on a multi-sensor array, characterized by: The method comprises: First step: obtaining an initial signal set by exposing a multi-sensor array to an odor sample; Second step: first, signal conditioning is performed on the initial signal set to generate a conditioned signal set; then, based on the temperature and humidity data of the environment in which the odor sample is located, dynamic error correction is performed on the conditioned signal set to generate an environment-compensated signal set; Third step: pattern recognition is performed on the environment-compensated signal set to obtain odor component-concentration relationship data of the odor sample.

2. The method according to claim 1, wherein: The method further comprises: Fourth step: converting the odor component-concentration relationship data into a standard format to generate a standardized data package.

3. The odor component analysis method based on a multi-sensor array according to claim 1 or 2, characterized in that: In the first step, the multi-sensor array comprises a plurality of electrochemical sensors, a plurality of optical sensors, and a plurality of semiconductor sensors.

4. The odor component analysis method based on a multi-sensor array according to claim 1 or 2, characterized in that: In the second step, the signal conditioning refers to sequentially performing amplification and denoising.

5. The odor component analysis method based on a multi-sensor array according to claim 1 or 2, characterized in that: In the second step, the dynamic error correction of the conditioned signal set based on the temperature and humidity data of the environment in which the odor sample is located to generate an environment-compensated signal set refers to: first, calling a pre-stored temperature and humidity-signal error mapping relationship, calculating the corresponding signal compensation amount based on the real-time collected temperature and humidity data of the environment in which the odor sample is located; then, using the signal compensation amount to correct the conditioned signal set to generate an environment-compensated signal set.

6. The method according to claim 5, wherein: The pre-stored temperature and humidity-signal error mapping relationship is obtained by the following method: Before obtaining the initial signal set, first, measure the response signal of the multi-sensor array to standard gas in a plurality of different temperature and humidity combinations of environments, and then based on the deviation of the response signal and the reference signal under different temperatures and humidities, establish the temperature and humidity-signal error mapping relationship through data fitting.

7. The odor component analysis method based on a multi-sensor array according to claim 1 or 2, characterized in that: In the third step, the pattern recognition of the environment-compensated signal set to obtain the odor component-concentration relationship data of the odor sample refers to: first, feature extraction is performed on the environment-compensated signal set to obtain a feature vector; then, the feature vector is input into a classification model to identify the category label of the odor component in the odor sample; then, based on the category label and the feature vector, a category odor component regression model is used to calculate the concentration value of the corresponding odor component; finally, the concentration values of each odor component are associated and combined with the corresponding category labels to generate the odor component-concentration relationship data of the odor sample.

8. The method according to claim 7, wherein: The classification model is obtained by the following method: First, a plurality of standard smell samples of known category labels are collected, and standard feature vectors corresponding to the standard smell samples are extracted, and then the machine learning algorithm is supervised trained with the standard feature vectors as input and the corresponding category labels as expected output, to form a mapping relationship from the feature vector to the category label, and finally a classification model is obtained; The category smell component regression model is obtained by the following way: A dedicated regression model is established for each category of smell components; wherein for each category of smell components, a plurality of standard samples of known concentration values are first collected, and the corresponding standard feature vectors are extracted, and then the regression algorithm is supervised trained with the feature vectors as input and the corresponding concentration values as expected output, to form a mapping relationship from the feature vector to the concentration value, and finally the dedicated regression model of the category smell component is obtained.

9. An apparatus for use in a method of resolving odor components based on a multi-sensor array according to claim 1 or 2, characterized in that: The device comprises a multi-sensor array (1), a data preprocessing unit (2), a power supply module (3); The data preprocessing unit (2) comprises a microprocessor (21), a signal conditioning circuit (22), and a temperature and humidity sensor (23); The input end of the signal conditioning circuit (22) is connected with the multi-sensor array (1), the output end of the signal conditioning circuit (22) is connected with the microprocessor (21), and the output end of the temperature and humidity sensor (23) is connected with the microprocessor (21); The power supply module (3) is connected with the multi-sensor array (1), the microprocessor (21), the signal conditioning circuit (22), and the temperature and humidity sensor (23) respectively.

10. The multi-sensor array based odor component resolution apparatus according to claim 9, wherein: The device further comprises a housing (5); The multi-sensor array (1), the data preprocessing unit (2), and the power supply module (3) are all arranged in the housing (5); the housing (5) is provided with an opening (51) at the front end, the sensing surface of the multi-sensor array (1) is arranged at the opening (51), and the opening (51) is provided with a protective cover (52).

Images