Nitrous oxide characteristic peak identification and background noise suppression method and system

Abstract

The invention relates to the technical field of gas detection, and discloses a laughing gas characteristic peak identification and background noise suppression method and a laughing gas characteristic peak identification and background noise suppression system. According to the method, background noise sample signals are subjected to deep analysis, real-time environment data prediction is fused to obtain noise reference signals, so that background noise is accurately tracked, and meanwhile, noise suppression processing is performed on original response signals based on an iterative adaptive filtering method of the noise reference signals. Not only is the signal-to-noise ratio of the signal remarkably improved, but also the dynamic evolution data of the signal in the noise reduction process can be obtained, and the dynamic interference stripping is performed on the signal with the high signal-to-noise ratio by combining the real-time environment data and the interferent response database, so that the known interferent influence and systematic influence can be actively and specifically eliminated, and the noise reduction accuracy is improved. The problems that a traditional method is insufficient in cross interference resistance and poor in characteristic peak recognition specificity are solved, and therefore the laughing gas characteristic peak recognition specificity and reliability are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of gas detection technology, and in particular to a method and system for identifying nitrous oxide characteristic peaks and suppressing background noise. Background Technology

[0002] Nitrous oxide, due to its unique physiological effects and ease of illegal acquisition and use, can negatively impact a user's judgment and ability to act, and may also disrupt public order and threaten public safety. Therefore, in the field of prohibited substance detection, the rapid and accurate identification of nitrous oxide content in gases under specific conditions through breath tests, chemical tests, and hair follicle tests has become a crucial means of ensuring public safety and order. Breath tests, in particular, place stringent requirements on the practicality and interference resistance of nitrous oxide detection: whether in mobile enforcement environments or densely populated public places, portable devices must be used for rapid detection responses, and the influence of interfering components in complex gases must be effectively eliminated to ensure that the test results provide a reliable basis for compliance determination.

[0003] Currently, in the process of breath testing, when there are multiple interfering components in the gas to be tested, such as common ethanol, acetone, hydrogen sulfide, or carbon dioxide, water vapor, volatile organic compounds commonly found in environmental monitoring, existing methods are unable to effectively distinguish the superimposed or similar response signals generated by these interfering substances and nitrous oxide on the sensor. Furthermore, gas detection relies on simple differential measurement or fixed threshold judgment, lacking the ability to accurately separate the characteristic signals of nitrous oxide from complex mixed signals. This makes it very easy for the system to misjudge the response of interfering substances as nitrous oxide, thus seriously affecting the accuracy and reliability of nitrous oxide identification. Summary of the Invention

[0004] The main objective of this invention is to provide a method and system for identifying nitrous oxide characteristic peaks and suppressing background noise, aiming to solve the technical problems in the prior art.

[0005] This invention proposes a method for identifying nitrous oxide characteristic peaks and suppressing background noise, comprising: The background noise sample signal and the original response signal corresponding to the current detected gas are obtained by the gas sensor, and real-time environmental data are obtained by the auxiliary sensor. Feature extraction is performed on the background noise sample signal to obtain a multidimensional feature vector, and the background noise sample signal is calculated based on the multidimensional feature vector and the real-time environmental data to generate a noise reference signal; The original response signal is subjected to noise suppression processing based on the noise reference signal to obtain a high signal-to-noise ratio signal and an iterative evolution sequence set; An interference response database is acquired, and the high signal-to-noise ratio signal is dynamically stripped of interference based on the real-time environmental data and the interference response database to obtain a purification response signal. An initial set of candidate peaks is obtained based on the purification response signal, and features are extracted from each candidate peak based on the iterative evolution sequence set to obtain the final set of candidate peaks, the static peak shape feature set, and the dynamic evolution feature set. Each candidate peak in the final candidate peak set is screened according to the static peak shape feature set and the dynamic evolution feature set to obtain the nitrous oxide feature peak.

[0006] This application also provides a system for identifying nitrous oxide characteristic peaks and suppressing background noise, including multiple modules, which are used to implement the steps of the above-mentioned method for identifying nitrous oxide characteristic peaks and suppressing background noise.

[0007] Preferably, the module includes multiple units, which are used to implement the steps of the above-described method for identifying nitrous oxide feature peaks and suppressing background noise.

[0008] The present invention also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the above-described method for identifying nitrous oxide feature peaks and suppressing background noise.

[0009] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the above-described method for identifying nitrous oxide feature peaks and suppressing background noise.

[0010] The beneficial effects of this invention are as follows: By deeply analyzing background noise sample signals and extracting their multidimensional feature vectors, and simultaneously fusing real-time environmental data to predict noise reference signals, this invention can effectively address the inherent and constantly changing nature of noise, thereby achieving accurate tracking of background noise. Furthermore, the iterative adaptive filtering method based on the noise reference signal performs noise suppression processing on the original response signal, significantly improving the signal-to-noise ratio (SNR) and allowing the acquisition of the dynamic response behavior of each component in the signal during the noise reduction process. Moreover, by combining real-time environmental data and an interference response database to dynamically strip high SNR signals from interference, it can proactively and specifically eliminate the systematic impact of known interferences and environmental fluctuations on the signal baseline and gas sensor sensitivity, thus improving signal purity from the source and solving the problems of baseline drift and sensitivity reduction caused by environmental changes in traditional methods. Finally, the method of cross-validating and intelligently judging each candidate peak in the final candidate peak set using both dynamic and static features significantly improves the specificity and reliability of nitrous oxide feature peak recognition. Attached Figure Description

[0011] Figure 1 This is a schematic diagram of a method flow according to an embodiment of the present invention.

[0012] Figure 2 This is a schematic diagram of the system structure according to an embodiment of the present invention.

[0013] Figure 3 This is a schematic diagram of the internal structure of a computer device according to an embodiment of this application.

[0014] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0015] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0016] like Figure 1 As shown, this application provides a method for identifying nitrous oxide characteristic peaks and suppressing background noise, including: S1. Obtain background noise sample signals and the original response signals corresponding to the gas being detected through a gas sensor, and obtain real-time environmental data through an auxiliary sensor; S2. Extract features from the background noise sample signal to obtain a multi-dimensional feature vector, and calculate the background noise sample signal based on the multi-dimensional feature vector and the real-time environmental data to generate a noise reference signal; S3. Based on the noise reference signal, perform noise suppression processing on the original response signal to obtain a high signal-to-noise ratio signal and an iterative evolution sequence set; S4. Obtain the interference response database, and perform dynamic interference stripping on the high signal-to-noise ratio signal based on the real-time environmental data and the interference response database to obtain the purification response signal; S5. Obtain an initial candidate peak set based on the purification response signal, and extract features from each candidate peak based on the iterative evolution sequence set to obtain the final candidate peak set, static peak shape feature set, and dynamic evolution feature set. S6. Based on the static peak shape feature set and the dynamic evolution feature set, each candidate peak in the final candidate peak set is screened to obtain the nitrous oxide feature peak.

[0017] As described in steps S1-S6 above, this invention actively samples background noise after the nitrous oxide detection device is started and before formal gas detection, obtaining a background noise sample signal. This background noise sample signal refers to the pure noise signal obtained by sampling the ambient air in the current detection environment through a gas sensor within a specific sampling window. This signal does not contain any response components generated by actively introduced nitrous oxide. Its signal structure fully characterizes the comprehensive background interference generated by the combined effects of sensor background electronic noise, ambient temperature and humidity fluctuations, and inherent non-target interfering gases in the ambient air under specific time and space conditions. The gas sensor in the nitrous oxide detection device used in this solution is a multi-channel sensor based on the non-dispersive infrared principle. Internally, it uses fixed filters to construct an N2O measurement channel, a reference channel, and an interference compensation channel, each working independently. Simultaneously, a continuously switched modulated infrared light source is used, allowing the sensor to receive continuous sinusoidal signals. The processor performs preliminary temperature drift and interference compensation by integrating the parameters of each channel. However, this... Fixed compensation mechanisms at the hardware level have inherent limitations. Traditional methods typically rely on fixed noise models or simple baseline subtraction to suppress noise in the signal. This approach, which attempts to use a fixed, pre-defined noise model or a single background sample to address noise suppression throughout the entire detection process, is inefficient and inaccurate. This invention, however, acquires background noise information from the current detection environment to provide a benchmark for predicting background noise during the detection period. Then, a gas sensor samples the gas being detected to obtain the raw response signal. This raw response signal is the mixed electrical signal directly output by the gas sensor when detecting the currently sampled gas, which may contain nitrous oxide, background noise, and various interfering substances. Simultaneously, real-time environmental data for the detection period is acquired based on the sensor. This real-time environmental data refers to a set of parameters reflecting the physical state of the current detection environment, including ambient temperature and humidity, synchronously collected by auxiliary sensors integrated into the device. This provides key parameters for subsequent signal correction and noise modeling. The characteristic peak of nitrous oxide specifically refers to the signal response peak in the output signal of a gas sensor, generated by the specific interaction between nitrous oxide molecules and the sensor, possessing specific positional and morphological characteristics. This peak exhibits a recognizable pattern in the time or frequency domain, distinguishing it from background noise and other interfering substances. It is the fundamental basis for achieving qualitative and quantitative identification and analysis of nitrous oxide. However, in practical gas detection applications, background noise is not a static, constant quantity, but a complex signal driven by multiple factors and continuously changing dynamically. Therefore, this invention performs in-depth analysis of background noise sample signals, extracts its multidimensional feature vectors, and integrates real-time environmental data to extrapolate the background noise sample signals, generating a noise reference signal. Here, the multidimensional feature vector refers to data that quantitatively describes the statistical and frequency domain characteristics of background noise, including skewness, kurtosis, etc. The noise reference signal refers to a signal sequence predicted based on the current noise characteristics and environmental conditions, capable of characterizing the continuous interference and random noise that exists in time and space during gas detection. This invention utilizes a method that collects background noise samples in real time before each detection and predicts the noise reference signal. This effectively addresses the inherent and ever-changing nature of noise, thereby achieving accurate tracking of background noise. Furthermore, since traditional filtering methods often lose signal details, this invention employs an iterative adaptive filtering method based on the noise reference signal to suppress noise in the original response signal. This process significantly improves the signal-to-noise ratio (SNR), resulting in a high SNR signal. Simultaneously, it captures the dynamic evolution of the signal from noisy to clean, yielding an iterative evolution sequence set. The high SNR signal refers to the signal obtained after iterative adaptive filtering of the original response signal to remove the noise reference signal. The iterative evolution sequence set refers to the complete historical data set of the intensity, contour, and stability parameters of each potential peak region in each iteration. This sequence set comprehensively characterizes the dynamic response behavior of each component in the signal during the noise reduction process, providing crucial evolutionary information for subsequent feature analysis. Next, an interference response database is acquired. This database, established experimentally beforehand, records the sensor response characteristics of common gaseous interferences under different environmental conditions. Its purpose is to provide standard comparison features for identifying and suppressing specific cross-interferences in real-time nitrous oxide detection. Combined with real-time environmental data, it dynamically corrects high signal-to-noise ratio signals. The core function of this step is to proactively and specifically eliminate the systematic impact of known interferences and environmental fluctuations on the signal baseline and gas sensor sensitivity. This transforms a high signal-to-noise ratio signal affected by multiple interferences, with the baseline drifting with ambient temperature and signal strength inaccurate due to sensor sensitivity fluctuations, into a purified response signal that eliminates the influence of known interferences, has a stable baseline after temperature compensation, and has highly consistent sensitivity after real-time correction. This improves the purity of the signal from the source and helps solve the problems of baseline drift and sensitivity reduction caused by environmental changes in traditional methods. This ensures that subsequent identification of nitrous oxide characteristic peaks can be performed on a stable and unified signal benchmark. Then, multiple candidate peaks are initially located from the purification response signal to obtain an initial candidate peak set. Candidate peaks refer to extreme regions whose signal amplitude is significantly higher than the local background baseline, initially identified by a preset peak detection algorithm. These regions are listed as analysis objects for further verification by the system because of the similarity between their channel (e.g., N2O measurement channel) and their initial peak position, peak height, and other morphological characteristics to nitrous oxide characteristic peaks. They are key intermediate states connecting the original signal and the final judgment result. Next, dynamic behavioral features are extracted and quantified for each candidate peak based on the iterative evolution sequence set. Multiple candidate peaks are then screened based on the extracted quantized features to obtain the final candidate peak set. Simultaneously, feature descriptions are performed on the candidate peaks based on the extracted features, resulting in a static peak shape feature set and a dynamic evolution feature set. The static peak shape feature set refers to the set of parameters extracted from the final state of the purification response signal that describe its stable morphology for each candidate peak, specifically including peak position, peak height, full width at half maximum (FWHM), and symmetry. The dynamic evolution feature set refers to the set of parameters extracted from the iterative evolution sequence set that describe the dynamic evolution of each candidate peak. This invention describes a set of parameters for the behavior pattern of nitrous oxide during the noise suppression iteration process, specifically including stability index, convergence coefficient, and morphological consistency index. Since genuine nitrous oxide characteristic peaks typically possess specific morphologies and exhibit stability during noise reduction, while interference peaks exhibit abnormal morphologies or disordered behavior, the final method utilizes both static and dynamic features to cross-validate and intelligently discriminate each candidate peak in the final candidate peak set, based on a static peak shape feature set and a dynamic evolution feature set, to obtain the nitrous oxide characteristic peak. The essence of this discrimination process lies in executing a dual-validation logic: first, the candidate peak must possess static features in the final signal that match the standard peak shape of nitrous oxide; second, throughout the entire iterative purification process before reaching the final state, the candidate peak must exhibit the stability and consistency expected of a genuine characteristic peak. Any candidate peak that fails to pass both of these stringent validations is discarded as a false peak. This invention significantly improves the specificity and reliability of background noise suppression and nitrous oxide characteristic peak identification by using this method of suppressing signal noise based on real-time background noise and by performing dual-feature identification and analysis on candidate peaks.

[0018] In one embodiment, step S2, which involves extracting features from the background noise sample signal to obtain a multidimensional feature vector, and then calculating the background noise sample signal based on the multidimensional feature vector and the real-time environmental data to generate a noise reference signal, includes: S21. Perform time-domain analysis on the background noise sample signal to obtain the skewness index and the excess kurtosis. S22. Perform a fast Fourier transform on the background noise sample signal to obtain the frequency band energy ratio and autocorrelation attenuation coefficient, and construct a multi-dimensional feature vector based on the skewness index, excess kurtosis, frequency band energy ratio and autocorrelation attenuation coefficient. S23. Obtain the reference temperature and sensor operating temperature of the gas sensor, and obtain the temperature and humidity joint disturbance index and sensor thermal state coefficient based on the reference temperature, sensor operating temperature and the real-time environmental data. S24. Based on the temperature and humidity combined disturbance index, the sensor thermal state coefficient and the multidimensional feature vector, the background noise sample signal is calculated in real time to obtain the noise reference signal corresponding to the detection of the gas being detected.

[0019] As described in steps S21-S24 above, this invention obtains a skewness index by calculating the ratio of the third central moment of the background noise sample signal sequence to the cube of its standard deviation. Next, to evaluate the sharpness and tail characteristics of the background noise sample signal distribution, the ratio of the fourth central moment of the background noise sample signal sequence to the fourth power of its standard deviation is calculated, and the standard kurtosis value of the Gaussian distribution (3) is subtracted from this result to obtain the super-kurtosis. Then, a fast Fourier transform is performed on the background noise sample signal to obtain the noise power spectral density. The noise power spectral density refers to the spectrum characterizing the power distribution of the background noise sample signal at different frequency components. Based on the noise power spectral density, the spectrum is divided into multiple preset frequency bands that reflect different noise sources. Then, the frequency band energy proportion of each frequency band is calculated based on the noise power spectral density, thereby revealing the frequency domain of the noise. Characteristics; simultaneously, to analyze the intrinsic structure and correlation of noise in the time dimension, a frequency domain method based on Fast Fourier Transform is used to calculate the autocorrelation function sequence of the background noise sample signal. This sequence fully reveals the degree of self-similarity of the signal under different time delays. On this basis, by performing linear or logarithmic interpolation search in the linear autocorrelation sequence starting from the maximum point (zero time delay), the time delay corresponding to the value decreasing to the maximum value 1 / e is accurately found, thus obtaining the autocorrelation attenuation coefficient. This value serves to provide a quantitative criterion for the type of noise. Therefore, the skewness index, excess kurtosis, autocorrelation attenuation coefficient, and the frequency band energy proportion of each frequency band obtained by the above method together constitute a multidimensional feature vector of the statistical characteristics and temporal structure of the background noise sample signal. This vector is used to accurately quantify and describe the essential attributes of the background noise. Next, the reference temperature and reference humidity preset during the sensor's factory or laboratory calibration phase are obtained, and the difference between the real-time ambient temperature and the reference temperature is calculated to obtain the real-time temperature difference. Simultaneously, the real-time ambient humidity is obtained from the ambient humidity data, and the difference between the real-time ambient humidity and the reference humidity is calculated to obtain the real-time humidity difference. Then, the formula "" is used to calculate the real-time humidity difference. "Calculate the combined temperature and humidity disturbance index, where, Indicates the combined temperature and humidity disturbance index. This represents the linear influence coefficient of temperature. Indicates the real-time temperature difference. This represents the linear influence coefficient of humidity. Indicates the real-time humidity difference. The coefficient representing the temperature and humidity coupling effect is based on the underlying logic that the sensor's noise floor is significantly influenced not only by the independent effects of temperature and humidity, but also by the interaction (coupling effect) between them. Therefore, this invention uses a polynomial model to approximate this complex nonlinear relationship. The formula uses the term "temperature and humidity coupling effect coefficient" to represent this effect. "The simulation measures the linear effects of individual changes in temperature and humidity on noise. For example, as temperature increases and the temperature difference widens, the conductivity of the sensor material may change, leading to a linear increase in the noise floor. Conversely, as humidity increases and the humidity difference widens, water molecules may be adsorbed, causing baseline drift. The formula..." "What is captured is the nonlinear coupling effect. When temperature and humidity change simultaneously, their impact on the sensor is far more than a simple superposition of their independent effects. The temperature and humidity coupling effect coefficient in the formula determines the strength and direction of this coupling effect. Subsequently, the instantaneous temperature difference between the sensor's operating temperature and the real-time ambient temperature is calculated. This value directly reflects the static degree to which the sensor deviates from the ideal thermal equilibrium state. At the same time, by monitoring the instantaneous temperature difference value sequence at continuous time points, its derivative is calculated to obtain the rate of change of temperature difference. This rate of change captures the intensity of the dynamic thermal disturbance that the sensor is experiencing. Then, a weighting coefficient obtained in advance through experimental calibration is obtained. This coefficient, as a dimensionless amplification factor, plays a core role in normalizing the rate of change of temperature difference with different physical meanings and dimensions to the same scale that can be linearly superimposed with the instantaneous temperature difference value, thereby quantifying the equivalent risk of dynamic thermal shock. Then, the product of the absolute value of the rate of change of temperature difference and the weighting coefficient is calculated, and the sum of this product and the instantaneous temperature difference value is calculated to obtain the sensor's thermal state coefficient. This value is a comprehensive indicator that fully reflects the current thermal stability of the sensor." Subsequently, the system first performs state initialization, mapping the multidimensional feature vector to a set of internal state parameters. This set of state parameters constitutes an initial estimate of the time-varying statistical characteristics of the background noise sample signal. Next, the system initiates a closed-loop prediction process: in each calculation cycle, the algorithm inputs the real-time acquired temperature and humidity combined disturbance index and sensor thermal state coefficient into a parameterized transaction equation. This equation describes the quantitative relationship between environmental conditions and noise characteristics. By solving this equation, a new set of predicted state parameter values ​​is dynamically generated. This process essentially uses the temperature and humidity combined disturbance index and sensor thermal state coefficient as control variables to continuously adjust the values ​​of the state parameters, thereby predicting the evolution of noise statistical characteristics caused by environmental fluctuations, and updating accordingly. After the predicted state parameters are obtained, the system uses a preset signal generator to synthesize the corresponding time-domain waveform, which serves as the noise reference signal for the current nitrous oxide detection moment. To ensure prediction accuracy, a recursive correction loop is also embedded in the system. When a new background noise sample signal is acquired, the system immediately extracts a new multidimensional feature vector from it and compares the new multidimensional feature vector with the predicted state parameters obtained through the above process to obtain the prediction deviation. Then, based on the prediction deviation, the internal state parameters of the system are optimally corrected using a recursive estimation algorithm (such as Kalman filtering). This method ensures that the noise reference signal can continuously track the non-stationary characteristics of the background noise caused by long-term environmental drift or changes in sensor performance.

[0020] In one embodiment, step S4, which involves performing noise suppression processing on the original response signal based on the noise reference signal to obtain a high signal-to-noise ratio signal and an iterative evolution sequence set, includes: S31. Using the noise reference signal as a reference input signal and the original response signal as the main input signal, two filtering processes are applied to it in parallel to obtain the first filtered signal and the second filtered signal. S32. Evaluate the first filtered signal and the second filtered signal respectively to obtain their respective residual coherence coefficients and background noise intensity values, and filter the first filtered signal and the second filtered signal based on the residual coherence coefficients and the background noise intensity values ​​to obtain the preferred filtered signal; S33. Extract the residual noise signal from the preferred filtered signal, and compare the residual noise signal with the noise reference signal corresponding to the current iteration round to obtain the statistical feature difference. S34. The noise reference signal is updated based on the statistical feature difference to generate a noise reference update signal, and the noise reference update signal is used as the reference input signal for the next iteration to repeat the iteration, thereby obtaining a high signal-to-noise ratio signal and an iterative evolution sequence set.

[0021] As described in steps S31-S34 above, this invention uses the original response signal as the main input signal and the noise reference signal as the reference input signal of the filter. Based on this, the following two filtering processes are performed on the original response signal in parallel: The LMS algorithm is used to quickly track changes in the environment to filter the original response signal, resulting in a first filtered signal; the RLS algorithm is used to achieve high-precision convergence in steady state to filter the original response signal, resulting in a second filtered signal. This competitive mechanism ensures that at least one algorithm can provide effective filtering in any interference scenario, thus preventing filtering failures caused by a single algorithm's mismatch with the environment. Following this, the first... The filtered signal and the second filtered signal undergo rigorous quantitative evaluation in two dimensions. Firstly, interference suppression is evaluated: the first and second filtered signals are standardized and aligned with the noise reference signal to ensure perfect time-sequence matching and eliminate potential DC bias. Then, the Welch average periodogram method is used to perform joint frequency domain analysis on the two aligned signals, calculating their respective discrete complex spectra. Based on each pair of discrete complex spectra, the cross-power spectrum is calculated, and the arithmetic mean of the cross-power spectrum of all segments is obtained, resulting in a smoothed and more stable average cross-power spectral density estimate. Next, the first filtered signal and the second filtered signal are calculated... The self-power spectra of the segments of the wave signal and the segments of the noise reference signal are calculated respectively, and the arithmetic mean of all self-power spectra is obtained to obtain the average self-power spectral density estimates for the first filtered signal and the noise reference signal. Subsequently, for each discrete frequency point, the product of the average self-power spectral density estimates for the first filtered signal and the noise reference signal is calculated, and the ratio of the square of the average cross-power spectral density estimate to the above product is further calculated to obtain the residual coherence coefficient. This value is used to measure whether there are still residual noise models in the candidate signal. If the residual coherence coefficient of the predicted interference component approaches 0, it indicates that the adaptive filter has successfully removed the known interference component modeled by the noise reference signal at that frequency point. The filtered signal at this frequency point only contains signals and random noise that are unrelated to the noise reference. If the residual coherence coefficient approaches 1, it indicates that the filtered signal at this frequency point still contains interference components that are highly similar to the noise reference signal. The adaptive filter has failed to effectively suppress the known interference at this frequency, and the filtering process is not effective in this frequency band. Similarly, the residual coherence coefficients of the second filtered signal and the noise reference signal at each discrete frequency point are calculated using the above method.Secondly, background noise level assessment is performed. Taking the assessment of the first filtered signal as an example, the specific steps are as follows: First, based on the iteration history and signal morphology self-recognition strategy, the background segment used for assessment is dynamically determined. Specifically, this is done in the following way: In the first iteration before the start of the iteration, the system uses the original response signal and selects a segment of the initial background segment that is stable, without obvious fluctuations, and pre-labeled as having no characteristic peaks. In subsequent iterations, signal segments near the same time point as the initial background segment reference area are preferentially used as candidate background segments. At the same time, in each iteration, the system analyzes the current first filtered signal in real time, calculates the local variance and first derivative of the signal within a sliding time window, and automatically identifies and marks those sufficiently long continuous signal segments whose local variance is consistently below the first preset threshold and whose absolute value of the first derivative is consistently below the second preset threshold as dynamic candidate background segments. Subsequently, the system caches the previous multiple iterations. The signal regions identified as dynamic candidate background segments during iteration are examined in the current iteration, and the regions corresponding to these historical dynamic candidate background segments in the current first filtered signal are checked. If a region can be marked as a dynamic candidate background segment in multiple consecutive iterations, then that region is used as a high-confidence background segment for evaluation in this round. After successfully identifying the high-confidence background segments, the signal sequences corresponding to these high-confidence background segments are precisely extracted from the first filtered signal, i.e., the candidate background noise signal sequences, and the variance of the candidate background noise signal sequences is calculated and used as the background noise intensity value. This value is used to quantify the fluctuation intensity of the background region of the filtered signal. The lower the background noise intensity value, the lower the background noise level after filtering, indicating that the overall signal-to-noise ratio of the candidate filtered signal is significantly improved, and the filter does not introduce significant signal distortion while suppressing interference. Similarly, the background noise intensity value of the second filtered signal is calculated using the above method. Then, the first and second filtered signals are weighted and comprehensively scored based on the residual coherence coefficient and background noise intensity. The filtered signal with the higher comprehensive score is selected as the preferred filtered signal for the current iteration. Subsequently, the system extracts a candidate background noise signal sequence from the preferred filtered signal selected in the current iteration and defines this sequence as the residual noise signal. Next, the residual noise signal is statistically compared with the noise reference signal used in this iteration, including calculating the difference in their energy ratio and probability distribution shape parameters within a specific frequency band, to obtain the statistical feature difference. If the statistical feature difference exceeds a preset tolerance, it proves that the initial background noise sample signal failed to fully represent the interference components in the actual environment, and an update mechanism is triggered: the residual noise signal is fused with the current noise reference signal, specifically by weighted averaging of its power spectrum and updating the predicted state parameters, thereby generating a more accurate one for the next iteration. The updated noise reference signal; simultaneously, the peak trajectory tracking module in the system begins to work in parallel. This module caches the preferred filtered signal in each iteration and performs cross-iteration peak correlation: by continuously and consistently detecting peaks in all cached preferred filtered signals from the first round to the current round, based on the continuity of the peak position, peak shape symmetry, and full width at half maximum of each detected candidate peak, it intelligently correlates potential peaks belonging to the same physical source in different iteration rounds, thereby establishing an iterative correlation sequence for each suspected nitrous oxide feature peak; finally, when the iteration process meets the preset convergence condition, the system terminates the iteration. At this time, the system not only outputs the preferred filtered signal of the final iteration round as the final generated high signal-to-noise ratio signal, but more importantly, the peak trajectory tracking module synchronously outputs an iterative evolution sequence set containing complete cross-iteration data records of all potential peaks. This set contains the iterative change characteristics corresponding to all successfully tracked potential feature peaks.

[0022] In one embodiment, step S4, which involves dynamically stripping the high signal-to-noise ratio signal from the real-time environmental data and the interference response database to obtain a purified response signal, includes: S41. Obtain environmental humidity data and real-time environmental temperature based on the real-time environmental data, and construct a real-time response vector of the currently detected gas based on the environmental humidity data and the high signal-to-noise ratio signal. S42. Construct a standard response vector for each type of interfering object based on the interfering object response database; S43. Obtain the interference confidence level between the currently detected gas and each interfering substance based on the real-time response vector and the standard response vector, and obtain the total cross-interference response based on the interference confidence level and the interfering substance response database; S44. Obtain the primary suppression signal based on the high signal-to-noise ratio signal and the total cross-interference response; S45. Obtain the baseline drift compensation model and sensitivity correction function, and obtain the real-time baseline drift amount and sensitivity correction factor based on the real-time ambient temperature, the baseline drift compensation model and the sensitivity correction function; S46. The primary suppression signal is corrected according to the real-time baseline drift and the sensitivity correction factor to obtain the purification response signal.

[0023] As described in steps S41-S46 above, the present invention extracts environmental humidity data and real-time environmental temperature from real-time environmental data, and performs a cyclic operation in each processing cycle based on the obtained high signal-to-noise ratio signal and environmental humidity data: First, within a time window of a specific duration, the arithmetic mean of all the quantized values ​​of the original electrical signals within that time window is calculated to obtain the real-time response amplitude. At the same time, a univariate linear regression analysis is performed with real-time environmental humidity as the independent variable and the quantized values ​​of the original electrical signals as the dependent variable. The obtained regression coefficient is used as the real-time response slope, and a real-time response vector is constructed based on the real-time response amplitude and the real-time response slope. The interference response database refers to a pre-calibrated dataset containing steady-state response values ​​of various gaseous interferences under different ambient humidity levels. The steady-state response value refers to the characteristic output signal of a certain interfering gas at a specific ambient humidity and concentration, after reaching equilibrium. Subsequently, humidity response calibration curves are constructed for each gaseous interference in the interference response database. Then, based on the humidity response calibration curves, the entire humidity range (e.g., 30%RH to 90%RH) is divided into multiple continuous humidity intervals at equal intervals or non-equal intervals according to the curve's trend. For each divided humidity interval, the interval response slope and interval response amplitude are obtained, and a standard response vector is constructed based on the interval response amplitude and interval response slope calculated for each interfering gas in all humidity intervals. Subsequently, the interval response slope and interval response amplitude corresponding to each interfering gas belonging to the same preset humidity range as the current average ambient humidity are extracted from the standard response vector, i.e., the current humidity feature vector. Then, the similarity between the current humidity feature vector and the real-time response vector is calculated to obtain the interference confidence. The interference confidence is a probabilistic index used to quantify the degree of certainty that the judgment of the presence of a specific interfering gas in the current environment is true. Then, each calculated interference confidence is compared with a preset threshold. If an interference confidence is greater than the preset threshold, it is determined that the specific interfering gas exists in the current detection environment. The sum of the steady-state response values ​​corresponding to all the identified interfering gases can be used as the total cross-interference response. Then, the difference between the quantized value of each original electrical signal in the high signal-to-noise ratio signal and the total cross-interference response is calculated to obtain the primary suppression signal to eliminate the cross-response signal generated by the specific interfering gas on the main sensor. If all interference confidences are not greater than the preset threshold, it is determined that there is no identifiable specific interfering gas in the current detection environment, and the high signal-to-noise ratio signal is directly used as the primary suppression signal. Since the baseline output of a gas sensor is determined by the physicochemical state of its sensitive material, and this state is simultaneously modulated by the nonlinear coupling effect of the external real-time ambient temperature and the internal operating temperature, this invention constructs a baseline drift compensation model with the real-time ambient temperature and the sensor operating temperature as joint inputs. This model effectively eliminates signal drift caused by purely physical temperature effects at its source. The specific construction method of the baseline drift compensation model is as follows: First, obtain the initial expression of the baseline drift compensation model: ,in, Indicates the baseline drift amount. , , , , , Indicates the regression coefficients to be determined. Indicates the real-time ambient temperature. The sensor's operating temperature is used as the basis for baseline calibration data. This is achieved by systematically changing and collecting the steady-state output values ​​of the sensor under different combinations of real-time ambient and sensor operating temperatures in a clean detection environment. Historical ambient and sensor operating temperatures are then acquired. Based on this baseline calibration data, the least squares method is used to fit the aforementioned binary quadratic polynomial regression model, solving for the undetermined regression coefficients. This yields a specific functional expression for the sensor that can accurately predict baseline drift under any temperature combination. Furthermore, since the sensor's sensitivity to the target gas strongly depends on its core operating temperature—as temperature directly determines the adsorption, desorption, and reaction kinetics of gas molecules on the sensitive material surface—this invention constructs a sensitive... A sensitivity correction function is used to ensure consistent sensor response sensitivity across different operating temperatures. The method for constructing the sensitivity correction function is as follows: First, at multiple different stable sensor operating temperatures, standard nitrous oxide of known concentrations is introduced into the sensor, and the steady-state response values ​​are recorded, with the sensitivity coefficient at each temperature point calculated. Next, a reference temperature is selected, and its corresponding sensitivity coefficient is used as a benchmark. The sensitivity correction factor for each of the remaining temperature points is calculated; this factor is defined as the ratio of the benchmark sensitivity to the current temperature sensitivity. Then, the least squares method is used to perform univariate linear or quadratic polynomial regression fitting on all the historical sensor operating temperatures and corresponding sensitivity correction factors obtained from this process. This yields a specific function expression with sensor operating temperature as the independent variable and sensitivity correction factor as the dependent variable. ,in, Indicates the sensitivity correction factor. Represents a constant term. Represents the regression coefficient of the linear term. Represents the regression coefficient of the quadratic term. The function representing the sensor's operating temperature is based on the underlying logic that the sensor's response sensitivity is determined by the surface chemical reaction kinetics of its sensitive material, and there is a clear dependence between the reaction rate and temperature. This thermodynamic effect can be accurately compensated through the aforementioned fitting function. Then, the sensor's operating temperature is obtained, and the real-time ambient temperature and the sensor's operating temperature are input into the baseline drift compensation model to obtain a predicted real-time baseline drift. The difference between the quantized value of each electrical signal in the primary suppression signal and the real-time baseline drift is calculated to obtain the baseline correction signal, thereby eliminating the coupling effect of physical temperature on the sensor's zero point. Subsequently, the sensor's operating temperature is substituted into the sensitivity correction function to obtain the sensitivity correction factor, and the ratio of the quantized value of each electrical signal in the baseline correction signal to the sensitivity correction factor is calculated to normalize the sensor's response sensitivity. Finally, a purified response signal with an abnormally stable baseline, calibrated sensitivity, and significantly weakened from the source due to specific cross-interference is output.

[0024] In one embodiment, step S5, which involves obtaining an initial candidate peak set based on the purification response signal and extracting features from each candidate peak based on the iterative evolution sequence set to obtain a final candidate peak set, a static peak shape feature set, and a dynamic evolution feature set, includes: S51. The continuous wavelet transform modulus maxima detection method is used to perform peak detection on the purification response signal to obtain the candidate peak time coordinates and candidate peak boundary intervals, and an initial candidate peak set is constructed based on the candidate peak time coordinates and candidate peak boundary intervals. S52. Associate and match each candidate peak in the initial candidate peak set with each candidate peak evolution sequence in the iterative evolution sequence set to obtain the complete evolution sequence corresponding to each candidate peak. S53. Extract the stability index corresponding to each candidate peak from the complete evolution sequence, and filter multiple candidate peaks according to the stability index to obtain the final candidate peak set. S54. Obtain the static peak shape feature set corresponding to each candidate peak according to the purification response signal, and obtain the dynamic evolution feature set corresponding to each candidate peak according to the iterative evolution sequence set.

[0025] As described in steps S51-S54 above, the present invention performs peak detection on the purification response signal using the continuous wavelet transform modulus maxima detection method to accurately locate the peak position and obtain the candidate peak time coordinates. Subsequently, based on the candidate peak time coordinates, the first derivative of the purification response signal is searched for zero-crossing points on both sides of the peak to determine the start and end points of each candidate peak, thereby obtaining the candidate peak boundary interval. Based on all candidate peaks detected in this way, as well as the candidate peak time coordinates and candidate peak boundary intervals corresponding to each candidate peak, an initial candidate peak set is constructed. Subsequently, each candidate peak in the initial candidate peak set is associated and matched with the iterative evolution sequence set. The specific matching algorithm is as follows: For a candidate peak in the initial candidate peak set, the system traverses each target peak evolution sequence in the iterative evolution sequence set. Each target peak evolution sequence contains the peak position sequence of the candidate peak in each iteration (from the first round to the final round). Next, the peak position time difference between the peak time point of the candidate peak to be matched and the final round position of the target peak position sequence is calculated. At the same time, the relative error between the full width at half maximum (FWHM) of the candidate peak to be matched and the full width at half maximum (FWHM) of the target peak evolution sequence in the final round record is calculated to obtain the peak shape consistency error. Then, the time alignment tolerance threshold and shape tolerance parameters are obtained. If the peak position time difference is less than the time alignment tolerance threshold and the peak shape consistency error is less than the shape consistency threshold, it is determined that the candidate peak to be matched is successfully associated with the target peak evolution sequence, and the complete target peak evolution sequence is assigned to the candidate peak to obtain the complete evolution sequence of the candidate peak. Next, from the complete evolution sequence associated with each initial candidate peak, the signal-to-noise ratio (SNR) iterative sequence, the final full width at half maximum (FWHM) sequence, and the coherence iterative sequence are extracted. Then, based on the SNR iterative sequence, a linear regression algorithm is used to fit a straight line with the iteration number as the independent variable and SNR as the dependent variable. The slope of this line is calculated and defined as the SNR growth rate, used to quantify the rate at which the signal strength of the candidate peak increases during iterative filtering. Next, the sample standard deviation of the final FWHM sequence is calculated to obtain the peak width volatility, which serves as an indicator of the convergence of the negative vectorized peak shape. The larger the peak width volatility, the more unstable the peak shape is at the end of the iteration, and the worse the convergence. Finally, based on the coherence iterative sequence, a linear regression algorithm is used... The method fits a straight line with the iteration number as the independent variable and the coherence coefficient as the dependent variable. The slope of this line is calculated and defined as the interference suppression rate. Since the effective filtering expected coherence decreases continuously, the ideal value of this value is negative, and its magnitude represents the rate at which known interference components are suppressed. Then, the stability indicators such as the signal-to-noise ratio growth rate, peak width fluctuation rate, and interference suppression rate are compared with the preset signal-to-noise ratio growth rate threshold, peak width fluctuation threshold, and interference suppression threshold, respectively. If any indicator does not meet the requirements, the candidate peak is removed from the initial candidate peak set as a false peak, and only those candidate peaks that meet the stability requirements of all indicators are confirmed as reliable candidate peaks, thus obtaining the final candidate peak set. Subsequently, for each candidate peak in the final candidate peak set, the original signal data within the boundary interval of the candidate peak is extracted from the purification response signal to obtain the candidate peak curve. Based on the candidate peak curve, the candidate peak time coordinate, candidate peak height, candidate peak full width at half maximum (FWHM), peak shape symmetry, and candidate peak area are obtained and used as the static peak shape feature set for each candidate peak. At the same time, based on the signal-to-noise ratio (SNR) iteration sequence, the SNR ratio ratio of the final round to the initial round is calculated to obtain the SNR enhancement gain. The negative value of the peak width fluctuation obtained in the previous steps is taken to obtain the peak shape convergence. The absolute value of the interference suppression rate obtained in the previous steps is taken to obtain the interference suppression index. The SNR enhancement gain, peak shape convergence, and interference suppression index are used as the dynamic evolution feature set for each candidate peak.

[0026] In one embodiment, step S6, which involves screening each candidate peak in the final candidate peak set based on the static peak shape feature set and the dynamic evolution feature set to obtain the nitrous oxide characteristic peak, includes: S61. Divide the static peak shape feature set into a core discrimination feature set and an interference-sensitive feature set; S62. Extract the standard features of nitrous oxide and the offset feature set of each interfering substance from the interfering substance response database, and obtain the morphological matching degree based on the standard features of nitrous oxide and the core identification feature set; S63. Each candidate peak in the final candidate peak set is screened according to the morphological matching degree to obtain multiple preliminary candidate peaks, and the interference matching index is obtained according to the interference sensitive feature set and the offset feature set. S64. Correct the morphological matching degree according to the interference matching index to obtain the corrected matching degree; S65. Obtain the dynamic evolution adjustment factor based on the dynamic evolution feature set, and obtain the comprehensive confidence probability value based on the dynamic evolution adjustment factor and the correction matching degree; S66. Obtain the confidence threshold, and filter multiple preliminary candidate peaks according to the confidence threshold and the comprehensive confidence probability value to obtain the nitrous oxide characteristic peak.

[0027] As described in steps S61-S66 above, this invention divides the static peak shape feature set into a core identification feature set and an interference-sensitive feature set. The core identification feature set includes the candidate peak time coordinates and peak shape symmetry, while the interference-sensitive feature set includes the candidate peak height, candidate peak full width at half maximum (FWHM), and candidate peak area. Then, for each candidate peak in the final candidate peak set, its core identification features are first matched with the nitrous oxide standard features: the standard time coordinates and standard peak shape symmetry corresponding to the nitrous oxide standard features are obtained. Next, the standardized deviations of the candidate peak time coordinates and peak shape symmetry relative to the standard time coordinates and standard peak shape symmetry are calculated, and then... A negative exponential weighted summation formula is used to calculate the morphological matching degree, which is used to characterize the overall similarity between the features of the candidate peak and the standard features of nitrous oxide. Then, the calculated morphological matching degree is compared with a preset first threshold: if the morphological matching degree is greater than the first threshold, the candidate peak is determined to have the basic morphology of nitrous oxide and is allowed to enter the next stage for more refined identification; if the morphological matching degree is not greater than the first threshold, the candidate peak is directly excluded from the subsequent process. By traversing each candidate peak in the final candidate peak set and filtering them, multiple preliminary candidate peaks are obtained, thereby optimizing the overall computational efficiency and strengthening the specificity of early screening. Then, the offset feature set for each type of interfering substance is extracted from the interfering substance response database. This offset feature set records typical values ​​of the gaseous interfering substance in features such as full width at half maximum (FWHM) and peak area, and quantifies the offset direction and magnitude of these feature values ​​relative to the standard features of nitrous oxide. Next, for each preliminarily identified candidate peak, the similarity between the interference-sensitive feature set of the preliminarily identified candidate peak and the offset feature set of each gaseous interfering substance is calculated. Then, based on the magnitude of the similarity, the N potential interfering substance types most similar to the current preliminarily identified candidate peak are identified. Following this, based on these N similarities and their inherent confidence levels in the interfering substance response database, which characterize the reliability of the interfering substance pattern, a weighted fusion is performed to calculate the interference matching index. This value is between 0 and 1; a higher value indicates that the preliminarily identified candidate peak is more likely to be a known gaseous interfering substance rather than nitrous oxide. Finally, the formula " "Calculate the corrected matching degree, where, Indicates the corrected matching degree. Morphological matching degree Indicates the interference matching index. The sharpening factor is represented by the formula's underlying logic, which uses the interference matching index between the initially identified candidate peak and known interfering objects as an objective suppression coefficient. This coefficient dynamically decays the morphological matching degree in a non-linear manner. The core suppression base in the formula is " This quantifies the confidence level of the preliminary candidate peak in terms of its difference from known gaseous interfering substances in terms of interference-sensitive features. When the interference matching index is zero, the suppression coefficient is 1, and the morphological matching degree is maintained. When the interference matching index increases, the suppression effect is enhanced. The key design lies in the sharpening factor (its value is greater than 1), which makes this suppression relationship exhibit non-linear sharpening characteristics: in the range of low interference matching index, the suppression change is gradual, avoiding over-suppression, but in the range of high interference matching index, the suppression effect will increase sharply. A small increase in the interference matching index will cause the matching degree to converge significantly to close to zero, thereby maximizing the elimination of false alarm signals caused by known cross-interference at the algorithm level. Next, each dynamic evolution feature in the dynamic evolution feature set undergoes independent threshold comparison and linear amplification. This processing follows a unified contribution quantification model. Specifically, taking the processing of signal-to-noise ratio enhancement gain as an example, the processing method is as follows: a minimum growth factor threshold is set, and the amplification is performed according to the formula " "Calculate the signal-to-noise ratio gain contribution value, where, This represents the signal-to-noise ratio gain contribution value. Indicates the magnification factor. This indicates the signal-to-noise ratio (SNR) improvement gain. This represents the minimum growth factor threshold. When the signal-to-noise ratio (SNR) gain exceeds this threshold, its contribution is quantified as "". When the signal-to-noise ratio enhancement gain is not greater than the minimum growth factor threshold, the contribution value remains "1". Similarly, the peak convergence degree and the interference suppression index are processed in parallel according to this model to generate the corresponding peak convergence contribution value and interference suppression contribution value, respectively. Then, the weighted geometric mean algorithm is used to combine the above three contribution values ​​into a single dynamic evolution adjustment factor. Finally, the product of the correction matching degree and the dynamic evolution adjustment factor is calculated to obtain the comprehensive confidence probability value. Finally, the overall confidence probability values ​​of all preliminarily identified candidate peaks are compared with a preset, strict confidence threshold. If the overall confidence probability values ​​of all preliminarily identified candidate peaks are not greater than the confidence threshold, it is determined that no valid nitrous oxide characteristic peaks were identified in this detection, and the "not detected" status is output. If the overall confidence probability value of one preliminarily identified candidate peak is greater than the confidence threshold, it is identified as a nitrous oxide characteristic peak. If the overall confidence probability values ​​of multiple preliminarily identified candidate peaks are greater than the confidence threshold, they are preliminarily marked as valid peaks, and a competitive exclusion mechanism is activated to handle complex scenarios where multiple valid peaks may exist and to prevent peak confusion caused by cross-interference: extracting from multiple valid peaks... The system combines the two valid peaks with the highest and second-highest comprehensive confidence probabilities and calculates the difference between their corresponding comprehensive confidence probabilities to obtain the probability score difference. When the probability score difference is greater than the preset minimum significance difference threshold, the valid peak with the highest comprehensive confidence probability is determined to be the nitrous oxide feature peak. When the probability score difference is not greater than the preset minimum significance difference threshold, it is determined that there is a suspected confusion between the two valid peaks. For such valid peak pairs with confusion, the system will no longer rely solely on the small difference in probability scores to make a decision, but will call the more discriminative dynamic evolutionary features as the ultimate arbiter, that is, the valid peak with the better evolutionary features (e.g., higher signal-to-noise ratio gain) will be taken as the nitrous oxide feature peak.

[0028] This application also provides a system for identifying nitrous oxide characteristic peaks and suppressing background noise, including: The data acquisition module is used to acquire background noise sample signals and the original response signals corresponding to the current detected gas through gas sensors, and to acquire real-time environmental data through auxiliary sensors. The noise acquisition module is used to extract features from the background noise sample signal to obtain a multi-dimensional feature vector, and to calculate the background noise sample signal based on the multi-dimensional feature vector and the real-time environmental data to generate a noise reference signal. A noise suppression module is used to perform noise suppression processing on the original response signal based on the noise reference signal to obtain a high signal-to-noise ratio signal and an iterative evolution sequence set. An interference stripping module is used to acquire an interference response database and dynamically strip the high signal-to-noise ratio signal from interference based on the real-time environmental data and the interference response database to obtain a purification response signal. The feature extraction module is used to obtain an initial candidate peak set based on the purification response signal, and to extract features from each candidate peak based on the iterative evolution sequence set to obtain a final candidate peak set, a static peak shape feature set, and a dynamic evolution feature set. The intelligent recognition module is used to screen each candidate peak in the final candidate peak set according to the static peak shape feature set and the dynamic evolution feature set to obtain the nitrous oxide characteristic peak.

[0029] In one embodiment, the noise suppression module includes: The filtering processing unit is used to take the noise reference signal as a reference input signal and the original response signal as the main input signal, and to perform two filtering processes on them in parallel to obtain a first filtered signal and a second filtered signal. The signal filtering unit is used to evaluate the first filtered signal and the second filtered signal respectively to obtain their respective residual coherence coefficients and background noise intensity values, and to filter the first filtered signal and the second filtered signal based on the residual coherence coefficients and the background noise intensity values ​​to obtain a preferred filtered signal; The difference evaluation unit is used to extract the residual noise signal from the preferred filtered signal and compare the residual noise signal with the noise reference signal corresponding to the current iteration round to obtain the statistical feature difference. The iterative update unit is used to update the noise reference signal based on the statistical feature difference, generate a noise reference update signal, and use the noise reference update signal as the reference input signal for the next iteration to repeat the iteration, thereby obtaining a high signal-to-noise ratio signal and an iterative evolution sequence set.

[0030] The present invention also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the above-described method for identifying nitrous oxide feature peaks and suppressing background noise.

[0031] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the above-described method for identifying nitrous oxide feature peaks and suppressing background noise.

[0032] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in this application and in the embodiments can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual-speed SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0033] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, apparatus, article, or method that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, apparatus, article, or method. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, apparatus, article, or method that includes that element.

[0034] The above description is merely a preferred embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A method for identifying nitrous oxide characteristic peaks and suppressing background noise, characterized in that, include: The background noise sample signal and the original response signal corresponding to the current detected gas are obtained by the gas sensor, and real-time environmental data are obtained by the auxiliary sensor. Feature extraction is performed on the background noise sample signal to obtain a multidimensional feature vector, and the background noise sample signal is calculated based on the multidimensional feature vector and the real-time environmental data to generate a noise reference signal; The original response signal is subjected to noise suppression processing based on the noise reference signal to obtain a high signal-to-noise ratio signal and an iterative evolution sequence set; An interference response database is acquired, and the high signal-to-noise ratio signal is dynamically stripped of interference based on the real-time environmental data and the interference response database to obtain a purification response signal. An initial set of candidate peaks is obtained based on the purification response signal, and features are extracted from each candidate peak based on the iterative evolution sequence set to obtain the final set of candidate peaks, the static peak shape feature set, and the dynamic evolution feature set. Each candidate peak in the final candidate peak set is screened according to the static peak shape feature set and the dynamic evolution feature set to obtain the nitrous oxide feature peak.

2. The method for identifying nitrous oxide characteristic peaks and suppressing background noise according to claim 1, characterized in that, The step of extracting features from the background noise sample signal to obtain a multi-dimensional feature vector, and then calculating the background noise sample signal based on the multi-dimensional feature vector and the real-time environmental data to generate a noise reference signal, includes: Time-domain analysis was performed on the background noise sample signal to obtain the skewness index and the excess kurtosis. A fast Fourier transform is performed on the background noise sample signal to obtain the frequency band energy ratio and autocorrelation attenuation coefficient, and a multidimensional feature vector is constructed based on the skewness index, excess kurtosis, frequency band energy ratio and autocorrelation attenuation coefficient. The reference temperature and operating temperature of the gas sensor are obtained, and the temperature and humidity joint disturbance index and the thermal state coefficient of the sensor are obtained based on the reference temperature, the operating temperature of the sensor and the real-time environmental data. Based on the temperature and humidity combined disturbance index, the sensor thermal state coefficient, and the multidimensional feature vector, the background noise sample signal is calculated in real time to obtain the noise reference signal corresponding to the detection of the gas being detected.

3. The method for identifying nitrous oxide characteristic peaks and suppressing background noise according to claim 1, characterized in that, The step of performing noise suppression processing on the original response signal based on the noise reference signal to obtain a high signal-to-noise ratio signal and an iterative evolution sequence set includes: The noise reference signal is used as the reference input signal, and the original response signal is used as the main input signal. Two filtering processes are applied to it in parallel to obtain the first filtered signal and the second filtered signal. The first filtered signal and the second filtered signal are evaluated respectively to obtain their respective residual coherence coefficients and background noise intensity values. Based on the residual coherence coefficients and the background noise intensity values, the first filtered signal and the second filtered signal are screened to obtain the preferred filtered signal. The residual noise signal is extracted from the preferred filtered signal, and the residual noise signal is compared with the noise reference signal corresponding to the current iteration round to obtain the statistical feature difference. The noise reference signal is updated based on the statistical feature difference to generate a noise reference update signal. The noise reference update signal is then used as the reference input signal for the next iteration to repeat the iteration, resulting in a high signal-to-noise ratio signal and an iterative evolution sequence set.

4. The method for identifying nitrous oxide characteristic peaks and suppressing background noise according to claim 1, characterized in that, The step of dynamically stripping interference from the high signal-to-noise ratio signal based on the real-time environmental data and the interference response database to obtain a purified response signal includes: The ambient humidity data and real-time ambient temperature are obtained based on the real-time environmental data, and a real-time response vector of the gas being detected is constructed based on the ambient humidity data and the high signal-to-noise ratio signal. Based on the interference response database, a standard response vector for each interference is constructed; The interference confidence level between the currently detected gas and each interfering substance is obtained based on the real-time response vector and the standard response vector, and the total cross-interference response is obtained based on the interference confidence level and the interfering substance response database. The primary suppression signal is obtained based on the high signal-to-noise ratio signal and the total cross-interference response. Obtain the baseline drift compensation model and sensitivity correction function, and obtain the real-time baseline drift amount and sensitivity correction factor based on the real-time ambient temperature, the baseline drift compensation model, and the sensitivity correction function; The primary suppression signal is corrected based on the real-time baseline drift and the sensitivity correction factor to obtain the purification response signal.

5. The method for identifying nitrous oxide characteristic peaks and suppressing background noise according to claim 1, characterized in that, The steps of obtaining an initial candidate peak set based on the purification response signal, and extracting features from each candidate peak based on the iterative evolution sequence set to obtain a final candidate peak set, a static peak shape feature set, and a dynamic evolution feature set include: The purification response signal is peaked using the continuous wavelet transform modulus maxima detection method to obtain the candidate peak time coordinates and candidate peak boundary intervals, and an initial candidate peak set is constructed based on the candidate peak time coordinates and candidate peak boundary intervals. Each candidate peak in the initial candidate peak set is associated and matched with each candidate peak evolution sequence in the iterative evolution sequence set to obtain the complete evolution sequence corresponding to each candidate peak. The stability index corresponding to each candidate peak is extracted from the complete evolution sequence, and multiple candidate peaks are screened according to the stability index to obtain the final candidate peak set. The static peak shape feature set corresponding to each candidate peak is obtained based on the purification response signal, and the dynamic evolution feature set corresponding to each candidate peak is obtained based on the iterative evolution sequence set.

6. The method for identifying nitrous oxide characteristic peaks and suppressing background noise according to claim 1, characterized in that, The step of screening each candidate peak in the final candidate peak set according to the static peak shape feature set and the dynamic evolution feature set to obtain the nitrous oxide characteristic peak includes: The static peak shape feature set is divided into a core discrimination feature set and an interference-sensitive feature set; Extract nitrous oxide standard features and offset feature sets for each interfering substance from the interfering substance response database, and obtain morphological matching degree based on the nitrous oxide standard features and the core identification feature set; Each candidate peak in the final candidate peak set is screened according to the morphological matching degree to obtain multiple preliminary candidate peaks, and the interference matching index is obtained according to the interference sensitive feature set and the offset feature set. The morphological matching degree is corrected based on the interference matching index to obtain the corrected matching degree; The dynamic evolution adjustment factor is obtained based on the dynamic evolution feature set, and the comprehensive confidence probability value is obtained based on the dynamic evolution adjustment factor and the correction matching degree. A confidence threshold is obtained, and multiple preliminary candidate peaks are screened based on the confidence threshold and the comprehensive confidence probability value to obtain the nitrous oxide characteristic peak.

7. A system for identifying nitrous oxide characteristic peaks and suppressing background noise, characterized in that, It includes multiple modules for implementing the steps of the method according to any one of claims 1 to 6.

8. The nitrous oxide characteristic peak recognition and background noise suppression system according to claim 7, characterized in that, It includes multiple units, which are used to implement the steps of the method according to any one of claims 1 to 6.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

Images