42 results about "Spectral centroid" patented technology

The invention discloses a glass ceramic thickness detection method and a related device, and the method comprises the steps: obtaining a vibration signal generated in a glass ceramic thinning and grinding process, and extracting a vibration dominant frequency, a frequency spectrum centroid frequency and a frequency band energy distribution ratio as thickness characteristic parameters through fast Fourier transform processing; three-dimensional position coordinates of the grinding head are obtained, space mapping and weighted average processing are conducted on the thickness characteristic parameters, and space distribution of the thickness characteristic parameters is obtained; performing process interference compensation and sliding window time sequence smoothing processing according to the grinding process parameters to obtain stable thickness characteristic distribution; based on the stable thickness feature distribution and the current grinding process, final thickness distribution is predicted, and a thickness qualification pre-judgment result is obtained; and extracting a time sequence change characteristic of the thickness characteristic parameter as a vibration evolution index, and performing comprehensive quality evaluation in combination with a thickness eligibility pre-judgment result. According to the scheme, the thickness can be detected in real time in the thinning and grinding process of the microcrystalline glass, and the production efficiency is improved.

The invention provides a remote monitoring method and system applied to children's watches, and relates to the technical field of children's watches, time axis synchronous alignment is carried out on collected multi-source heterogeneous data through an interpolation method, data time sequence chaos caused by sampling rate differences is prevented, and then the data time sequence chaos caused by sampling rate differences is prevented. Through dynamic coupling calculation of a light intensity mutation gradient and a heart rate stress coefficient, the cooperative influence of environment change on the physiological state of the child is quantified in real time, and the physiological state of the child is determined by further combining power spectrum centroid frequency and physiological phase difference analysis of the motion rhythm and logarithmic correlation of acoustic envelope frequency and GPS speed. In addition, an acoustic propagation delay compensation mechanism is introduced, a two-factor dynamic equation model is constructed through a time delay cross term, and a dynamic risk threshold generation mechanism based on a noise energy ratio is combined, so that the dynamic risk threshold of the motion intensity, the rhythm characteristics and the acoustic scene is effectively distinguished. Effective remote monitoring on child safety in different scenes is realized.

The invention belongs to the technical field of audio and video processing, and discloses an audio and video stream multi-mode real-time anomaly detection method and system based on intelligent feature tracking, and the method comprises the steps: carrying out the gray-scale map conversion, feature point extraction and global alignment, precise matching of a current frame and a historical frame, and judgment of abnormal feature points through the combination of feature point displacement and local gray scale difference; and subtle changes in the video can be effectively captured. Based on a dynamic threshold value, video abnormity is judged by combining the number of abnormal, disappearing and newly-added feature points, and the adaptability and accuracy of detection are enhanced. Meanwhile, sampling parameters are set, framing and STFT processing are carried out, a Mel filter bank is used for extracting a logarithmic Mel spectrogram, and through multi-index comparison of the energy ratio, the spectrum center difference value and the Frobenius norm difference, the audio abnormity is comprehensively analyzed. Audio and video are combined in multiple modes, comprehensive judgment is carried out from different dimensions, the comprehensiveness and reliability of anomaly detection are improved, and the method is suitable for scenes such as real-time monitoring.

The invention belongs to the field of power equipment state monitoring, and relates to a generator partial discharge on-line monitoring method, system, equipment and medium, and the method comprises the following steps: obtaining four types of original signals of a generator; extracting a partial discharge feature significant signal layer from the original signal by using an adaptive hierarchical denoising algorithm; constructing a feature quantity library based on the partial discharge feature significant signal layer, and extracting the frequency spectrum gravity center and wavelet energy entropy of each type of original signals; performing fuzzy logic weighting on the frequency spectrum gravity center and the wavelet energy entropy of each type of original signals to obtain a primary fusion result of each type of original signals; performing deep fusion on the primary fusion result of each type of original signals by adopting an LSTM neural network to obtain a partial discharge primary judgment result; and obtaining the score of the partial discharge preliminary judgment result based on the abnormal degree score of the mahalanobis distance, and if the score is greater than a decision threshold, determining that partial discharge occurs. The partial discharge of the generator can be quickly and effectively judged, and field operation and maintenance personnel can respond to the partial discharge conveniently.

The invention relates to the field of electrical variable measurement, in particular to an insulating layer compactness detection method for a mineral cable, which comprises the following steps: firstly, acquiring a time domain reflection signal of the cable, and carrying out denoising pretreatment by utilizing synchronous average and wavelet transform; and then, carrying out spectrum analysis on the discrete signal segments, calculating a spectrum centroid offset rate to determine high-frequency energy attenuation, and constructing a dynamic compensation gain coefficient according to the high-frequency energy attenuation. And then, improving a traditional layer stripping algorithm by using the gain coefficient, eliminating calculation deviation caused by transmission loss, and accurately reconstructing characteristic impedance distribution along the cable. And finally, by comparing the deviation between the reconstructed impedance and the nominal impedance, positioning and judging the abnormities of the compactness of the insulating layer of the cable and the eccentricity defect of the cable core are realized.

The invention belongs to the technical field of rotating equipment fault diagnosis, and particularly relates to a voiceprint-based rotating equipment fault open set diagnosis method, system and device. The method comprises the following steps: acquiring a voiceprint signal of a rotating device; preprocessing the voiceprint signal, extracting a Mel-frequency cepstrum coefficient and a sound spectrum centroid feature based on the preprocessed voiceprint signal, and fusing the Mel-frequency cepstrum coefficient and the sound spectrum centroid feature to obtain a fused feature; mapping the fusion feature to a potential space through an encoder, decoupling the fusion feature into a classification variable and a noise variable, and reconstructing the fusion feature through a decoder to obtain a reconstructed feature; calculating a prototype center of each fault category based on the classification variables; training a variational auto-encoder model based on the total loss function; and measuring the distance between the to-be-detected sample and the center of the known category prototype through the Euclidean distance and the dot product similarity, and judging whether the to-be-detected sample is an unknown fault category in combination with a preset threshold value. According to the invention, fault diagnosis of the rotating equipment is realized.

The invention discloses a slight overdischarge state identification method and equipment based on an ultrasonic phased array technology, and belongs to the technical field of lithium batteries. The method comprises the following steps: establishing a lithium battery ultrasonic phased array detection system, and collecting full-matrix ultrasonic data in charge and discharge cycles; carrying out superposition synthesis on the full-matrix ultrasonic data, and extracting frequency spectrum centroid characteristics of a synthetic wave to reflect frequency center deviation of an ultrasonic signal so as to realize characterization of a slight overdischarge behavior; in combination with spatial position information of phased array elements, mapping the spectrum centroid features to a physical space, constructing a spatially resolved spectrum centroid map, and realizing collaborative representation of spatial domain and frequency domain information; and performing Gabor multi-scale texture enhancement on the frequency spectrum centroid diagram, extracting spatial features representing local frequency anomaly, and inputting the spatial features into a KNN identification model, thereby realizing identification of the slight overdischarge state of the lithium battery. According to the invention, rapid and nondestructive detection of early-stage electric abuse damage of the battery can be realized.

The application provides a GH4169 forging grain structure ultrasonic evaluation method positioned along sound path zoning, extracts frequency spectrum centroid and high frequency / low frequency energy ratio as frequency domain characteristic parameters through wavelet packet decomposition, fully utilizes the physical law of high frequency scattering enhancement of coarse grain area, and has higher organization state distinguishing sensitivity compared with traditional time domain statistics; the application also provides a double threshold value scoring positioning method based on statistical baseline and an unsupervised positioning method based on K-means clustering, the overlapping identification rate of the two methods in coarse grain dominant area reaches more than 90%, and the two methods can be flexibly selected or verified according to the presence or absence of prior knowledge, and the reliability of the detection result is improved; the sliding window density criterion in the application realizes the judgment from single point anomaly to area scale by investigating the spatial aggregation of abnormal windows along the depth direction, effectively avoids false alarm caused by isolated noise points, and significantly improves the robustness of the detection result.

The invention provides a sound wave optical cable positioning method and system, and relates to the technical field of deep learning, and the method comprises the steps: uniformly arranging vibration sensors on a sensing optical cable, demodulating a back scattering light signal to obtain a differential phase signal, and constructing a self-correlation loss function. Preprocessing sound source data, extracting an audio frequency semantic feature vector, and calculating a dot product of a frequency spectrum centroid and frequency spectrum flatness to obtain a frequency spectrum loss function; and a phase change loss function is obtained by combining the ratio and the refractive index of the strained optical fiber and the non-strained optical fiber. And weighting the three loss functions to form a fusion loss function. And constructing a sound wave positioning model, and training the model to be optimal based on the fusion loss function. And finally, segmenting and enhancing audio data, and inputting the audio data into the trained model to obtain a positioning result. According to the method, the positioning model is trained by fusing multiple different types of loss functions, so that the positioning precision of sound waves in the optical cable is improved.

The invention relates to a method and a device for identifying a void defect of a steel plate concrete structure. The method comprises the following steps: acquiring a to-be-identified audio signal of the steel plate concrete structure acquired through excitation conditions; extracting feature information on each frame of the preprocessed audio signal, wherein the feature information at least comprises a Mel-frequency cepstrum coefficient, root mean square energy and two items of a spectral centroid statistical mean value and a standard deviation; the extracted frame feature information is input into a void defect recognition model, void confidence of each frame is obtained, and the void defect recognition model is obtained through training by utilizing a marked audio sample training set based on a support vector machine; in response to the fact that the void confidence coefficient corresponding to each frame of audio is higher than a confidence threshold value, judging that the corresponding frame is a void frame; and taking a judgment result of the majority frames as a final void defect identification result of the audio signal to be identified. The problems that a steel plate concrete structure is low in small-size void defect sensitivity and is greatly interfered by the thickness of a steel plate and an internal complex structure are solved.

The invention relates to the technical field of information security, and discloses a multi-point recording interference method and system, and the method comprises the steps: obtaining environment information, and determining a basic interference volume according to the environment information; acquiring a sound source-node distance vector and sound field uniformity, and converting the sound source-node distance vector and the sound field uniformity into a spatial distribution coefficient; acquiring a voice initial instantaneous slope and an interference-voice cross-correlation function peak value, and converting the slope and the peak value into a dynamic response coefficient; receiving a target voice, obtaining the spectral characteristics and the voice spectrum centroid of the target voice, and converting the spectral characteristics and the voice spectrum centroid into a spectral efficiency coefficient; reading the basic interference volume, the spatial distribution coefficient, the dynamic response coefficient and the spectral efficiency coefficient, and generating an adjustment instruction; according to the invention, through collaborative analysis of the environment reference coefficient, the spatial distribution coefficient, the dynamic response coefficient and the spectrum efficiency coefficient, multi-dimensional adaptive matching of the interference volume and the environment state, the spatial layout, the voice dynamic characteristic and the spectrum characteristic is realized, and the flexibility is extremely high.

The application discloses a kind of intelligent digital media content sentiment analysis method and system, content includes multi-source data acquisition, multi-level feature extraction, cross-modal fusion sentiment analysis modeling and intelligent analysis.The application relates to the technical field of media content sentiment analysis, specifically refers to a kind of intelligent digital media content sentiment analysis method and system, the scheme is in text passage calculation sentiment embedding offset and dynamically updates multi-center vector, in visual passage through space aggregation convolution activation measures emotional intensity, in audio passage using spectral centroid third central moment intensity is measured;At the same time, the key word intensity index and the modification mode coefficient are combined to realize the enhancement of middle-level semantics, and the global correlation of multi-modal nodes is constructed through the hypergraph structure, the context gate dynamic updating hyperedge weight is innovatively introduced, the graph attention aggregation feature is used, the time sequence self-attention and graph convolution are combined for spatiotemporal collaborative coding, which improves the accuracy, robustness and real-time adaptive ability of sentiment recognition.

The application discloses an intelligent noise monitoring system, comprising: a distributed sampling probe array installed on the bottom of a vehicle, a carriage and a skirt plate; a damping protection unit; a data acquisition and preprocessing unit; a vehicle-mounted multi-source information fusion module; an intelligent analysis and processing unit containing feature extraction, deep learning classification, anomaly detection and comprehensive evaluation modules; a remote operation and maintenance and self-calibration unit; a data storage and communication unit, which collects multi-channel noise and synchronizes the vehicle state; extracts MFCC, short-time energy, spectral centroid and formant features; inputs the feature matrix into a CRNN model to identify noise event types; calculates an anomaly score using an isolation forest and LOF fusion model; and judges the anomaly in combination with the vehicle state, classification probability and anomaly score. The application solves the problems that the existing system cannot adapt to the subway environment, has low recognition accuracy and poor operation and maintenance efficiency, and realizes precise monitoring and intelligent operation and maintenance.

The application discloses a bridge wind-induced vibration adaptive early warning method and system, the method acquires bridge acceleration time history data and synchronous wind speed data, and divides and analyzes samples according to a preset time interval; the acceleration time history is pretreated, and the response is decomposed into multiple intrinsic mode functions by using variational mode decomposition; the instantaneous frequency, instantaneous energy and instantaneous frequency entropy of each mode are calculated, the dominant mode of wind-induced vibration is screened according to the instantaneous frequency entropy and central frequency, and the wind-induced vibration time history is reconstructed; characteristics such as root mean square, peak factor, spectral centroid and the like are extracted from the wind-induced vibration time history, and an average wind speed is fused to construct a comprehensive early warning index; a peak threshold extreme value model is used to dynamically update the early warning threshold, and a Bayesian online change point detection is combined to identify risk mutations, and finally, graded early warning information is output. The application can inhibit the influence of vehicle vibration and other operation interference on wind-induced vibration risk identification, and improve the early warning capability of wind-induced risks such as vortex-induced vibration and buffeting.

This invention relates to a method for identifying weathered rock strata interfaces and adaptively controlling excavation parameters based on vibration spectrum analysis, belonging to the field of intelligent excavator construction technology. It addresses the problems of relying on manual experience for soil-rock interface identification during weathered rock strata trench excavation, low efficiency of mechanical switching, and slope vibration risks. The method collects vibration acceleration signals using a triaxial accelerometer, extracts frequency domain feature vectors including the dominant frequency amplitude, high-frequency band energy proportion, spectral centroid, and kurtosis value, and inputs these vectors into a multi-class support vector machine model to output category labels. When the category label changes from "cohesive silt" to "strongly weathered gneiss," the input current to the proportional solenoid valve of the excavator's hydraulic main pump is reduced, and a breaker hammer warning signal is activated. When the category label is "moderately weathered gneiss with interbedded boulders" and the high-frequency band energy proportion and kurtosis value exceed the safe range, the pilot oil pressure of the breaker hammer is limited. This method enables real-time identification of the soil-rock interface and adaptive control of excavation parameters.

The invention discloses a heading machine cutting head fault diagnosis method based on a frequency gating tree topology network. The specific implementation process is as follows: channel rearrangement is guided by using the spectrum centroid of a channel, accurate feature decoupling is performed on the channel by using anisotropic convolution in combination with the energy distribution characteristics of high and low frequency signals, and semantic dynamic generation of a sample specific weight is captured in real time through a multi-branch convolution gating mechanism. The tree topology structure brings multi-view key features such as accurate coordinates, a long-range structure and global context to the trunk network, and in order to make full use of the key features, a tree structure-oriented distillation strategy is provided. And mapping the injected multi-view key features and the output features of the backbone network to the same semantic metric space by using a parameter sharing classification mechanism, and obtaining a semantic soft mask by means of confidence transformation. Experiments show that by means of an intelligent feature extraction mechanism, the workload of fault diagnosis can be effectively reduced, and meanwhile the diagnosis efficiency is remarkably improved.

This invention relates to the field of industrial automation control technology, specifically disclosing a fault diagnosis method and system for control valves. The method includes: acquiring operating data of the control valve, including pressure data, temperature data, and vibration acoustic data; extracting features from the vibration acoustic data, obtaining features including root mean square value, peak value, peak factor, spectral centroid, energy ratio of a specific frequency band, and main direction of the sound source; wherein, based on a dynamic adaptive beamforming algorithm, spatial filtering and anti-interference mechanisms are introduced to calculate the main direction of the sound source; inputting a fault diagnosis model to calculate the fault type and fault probability; when a fault occurs, adjusting the valve according to the operating data. This invention can accurately identify various control valve faults, and the classification accuracy of cavitation faults reaches over 92%, making it particularly suitable for cavitation fault identification of electro-hydraulic linkage control valves under harsh operating conditions such as high pressure, large diameter, large pressure difference, and high precision regulation.

The invention discloses a pump unit cavitation fault diagnosis method based on a multi-modal physical information neural network, and the method comprises the steps: calculating vibration logarithm total energy and a frequency spectrum centroid based on vibration signal data, calculating a current standard deviation based on current signal data, constructing and training a teacher model composed of three parallel independent sub-networks, and carrying out the fault diagnosis of a pump unit cavitation fault. A student model is constructed and trained, the student model comprises a shared vibration encoder, a shared current encoder, a feature fusion layer, a fault classification head and a physical regression head, the weighted sum of classification loss, physical consistency loss and physical evolution trend constraint loss is adopted as a total loss function, and the student model is trained in a knowledge distillation mode. And forcing the model to learn a characteristic manifold conforming to a cavitation physical evolution rule. According to the method, the physical mechanism is embedded into the neural network through the PINN architecture, and the problems that a pure data driving model lacks physical interpretability and is weak in generalization ability are solved.

The invention relates to the technical field of digital audio processing, and discloses a music generation method, device and equipment based on cat call characteristics and a storage medium, and the method comprises the steps: extracting cat call sound, calculating a fundamental frequency track, a harmonic-to-noise ratio mark, formant characteristics, transient intensity and a frequency spectrum centroid in the cat call sound, forming a multi-dimensional characteristic vector, and obtaining a multi-dimensional characteristic vector; the problem that the cat call core acoustic information cannot be accurately captured by applying a human singing feature extraction method in an existing scheme is solved. Re-sampling the multi-dimensional feature vector to a beat resolution, and generating a control parameter for beat alignment, thereby solving the problem that the features are difficult to match with music beats due to the non-periodic characteristics of cat call; and mapping the control parameters into rhythm grids, chord labels and timbre vectors, and inputting the rhythm grids, chord labels and timbre vectors to a control interface of the music generation model to output music segments synchronized with the cat call features, and compared with the prior art, the problem that the cat call features cannot be effectively integrated into music generation due to the lack of an adaptive interface in an existing model is solved.

Provided is an optical fiber sensor that is capable of highly accurately detecting an abnormality in a structure from vibration information. This optical fiber sensor includes an optical fiber that is laid in the vicinity of a structure, a light source for introducing pulsed light of a specific period into the optical fiber, and an optical sensor for detecting return light that has been obtained as a result of the introduction of the pulsed light into the optical fiber. The structure is determined to have an abnormality if the spectral centroid of vibration information exceeds a threshold.

The application discloses a pump unit cavitation fault diagnosis method based on a multi-modal physical information neural network, comprising the following steps: calculating vibration logarithmic total energy and spectrum centroid based on vibration signal data, calculating current standard deviation based on current signal data, constructing and training a teacher model composed of three parallel independent sub-networks, constructing and training a student model, wherein the student model comprises a shared vibration encoder, a shared current encoder, a feature fusion layer, a fault classification head and a physical regression head, adopting a weighted sum of a classification loss, a physical consistency loss and a physical evolution trend constraint loss as a total loss function, training the student model through a knowledge distillation mode, and forcing the model to learn a feature manifold conforming to a cavitation physical evolution law. The application embeds a physical mechanism into a neural network through a PINN architecture, and solves the problems of lack of physical interpretability and weak generalization ability of a pure data-driven model.

This invention discloses a product design strategy generation method based on environmental feature extraction, belonging to the field of intelligent product design technology. The steps are as follows: constructing a multimodal dataset containing ambient sound, scene images, and odor data; extracting root mean square energy, zero-crossing rate, spectral centroid, and spectral flux features from the ambient sound; extracting explicit visual features from the scene images and implicit style features through an improved Swin-Transformer model; extracting intensity, pleasantness, and familiarity features from the odor data; performing semantic mapping through Mamdani fuzzy inference to generate a design guidance parameter vector, including sensory stimulation, morphological simplicity, emotional affinity, technological feel, interaction rhythm, stability, sense of security, and environmental integration; and generating a product design strategy based on this vector. This invention automatically transforms environmental perception into a product design strategy, improving the objectivity and efficiency of design decisions.

The application discloses a cross-domain migration method and device for intelligent diagnosis, and belongs to the technical field of signal processing and intelligent diagnosis. The cumulative energy frequency and the spectrum centroid distribution of a vibration signal are calculated, a unified sampling frequency is selected based on time-frequency feature difference, and feature set unification and standardization preprocessing are carried out; a subspace alignment algorithm with maximum mean difference constraint is adopted to map source domain and target domain features to a shared low-dimensional subspace, so that the distribution difference is reduced; a multi-task learning model with fusion credibility decision is further constructed, and through dynamic weight distribution and credibility weighted fusion, the final fault diagnosis of the target domain sample is realized. The application improves the accuracy and adaptability of fault diagnosis under cross-condition and cross-equipment conditions.

The application discloses a speech structure sensitive area constraint enhancement method and system for a coal mine scene, and the method comprises the following steps: obtaining an original amplitude spectrum and a high-order statistical quantity feature vector corresponding to a noisy speech signal, and performing splicing processing on the two to obtain a composite tensor; time dimension features and frequency dimension features are extracted based on the composite tensor, a preliminary enhanced amplitude spectrum is further obtained, and a joint feature tensor comprising a spectral centroid change rate, spectral flatness, an inter-frame correlation coefficient and a medium frequency energy proportion is constructed; a sensitive probability graph is constructed based on the joint feature tensor, which is isomorphic with the original amplitude spectrum; a suppression penalty term is constructed according to the sensitive probability graph, the preliminary enhanced amplitude spectrum and a pure speech signal amplitude spectrum; and a preset speech deep learning model is optimized and trained by using the suppression penalty term. The application can consider both the suppression of non-stationary transient noise and the preservation of the sensitive area of the speech structure, and solves the problem of the decline of speech quality and communication intelligibility under extreme working conditions.

The application relates to the field of device life prediction, and provides a contact wear intelligent evaluation method, system, device, product and medium, which comprises the following steps: measuring contact closing vibration data through a vibration sensor and measuring contact closing temperature data through a temperature sensor; performing self-calibration vibration data collection and frequency response function identification to obtain a frequency response function, and obtaining interface coupling characteristic parameters according to the frequency response function; compensating the contact closing vibration data, correcting the spectral centroid, and obtaining compensated vibration parameters; performing short-time Fourier transform and modulus conversion to obtain a two-dimensional sound spectrogram, constructing a convolutional neural network, inputting the two-dimensional sound spectrogram into the convolutional neural network, and obtaining a first wear amount; performing band-pass filtering to obtain high-frequency vibration energy, obtaining a second wear amount according to the high-frequency vibration energy, performing weighted summation to obtain a contact wear amount, evaluating contact residual life through the contact wear amount, and completing evaluation of the contact wear.

The invention discloses an intelligent digital media content sentiment analysis method and system. The intelligent digital media content sentiment analysis method comprises the steps of multi-source data collection, multistage feature extraction, cross-modal fusion sentiment analysis modeling and intelligent analysis. The invention relates to the technical field of media content sentiment analysis, in particular to an intelligent digital media content sentiment analysis method and system.According to the scheme, sentiment embedding offset is calculated in a text path, a multi-center vector is dynamically updated, sentiment intensity is measured through spatial aggregation convolution activation in a visual path, and sentiment analysis is achieved. Measuring the intense degree in an audio channel by utilizing a third-order central moment of a frequency spectrum centroid; meanwhile, middle-layer semantic enhancement is achieved in combination with a keyword intensity index and a modification mode coefficient, multi-modal node global association is constructed through a hypergraph structure, context gating is innovatively introduced to dynamically update hyperedge weights, graph attention aggregation features are adopted, time sequence self-attention and graph convolution are combined to conduct space-time collaborative coding, and the time sequence self-attention and graph convolution are combined to achieve time-space collaborative coding. And the accuracy, robustness and real-time adaptive capability of emotion recognition are improved.

This invention relates to the field of physiological signal analysis technology, specifically to an automatic analysis method and system for electroskin activity (ESA) signals based on multi-domain features. The method includes the following steps: acquiring ESA signals and correcting abnormal fluctuations within a fixed time window to obtain stable and continuous time-series data; filtering and downsampling the corrected signals to characterize the time-domain stability; analyzing signal change trends using moving averages and extracting low-frequency evolution and differential changes to form a multi-component joint characterization; iteratively calculating the multi-component power distribution and spectral centroid features from the frequency domain; statistically judging the frequency domain consistency of the multi-components; and fusing time-domain stability information to form a multi-domain feature analysis result. In this invention, robust statistical correction of abnormal fluctuations and characterization of signal stability ensures that the ESA time series remains consistent under noise drift. The fusion of low-frequency and differential changes constructs a multi-component characterization, and combined with frequency domain consistency analysis, achieves time-frequency verification, improving the stability and comparability of the results.

The invention discloses a mechanical valve closing detection method and device based on audio feature analysis, equipment and a medium, and relates to the technical field of industry, and the method comprises the steps: preprocessing an audio signal in an operation process of a field mechanical valve, segmenting the obtained processed audio signal, and obtaining a segmented audio signal; determining a sound energy value and a zero-crossing rate of each audio segment obtained by segmentation, performing short-time Fourier transform on the processed audio signal, and determining a Mel frequency cepstral coefficient by using an obtained frequency spectrum and a Mel filter bank; determining a spectral centroid corresponding to the frequency spectrum, performing multi-scale analysis on the processed audio signal based on a wavelet transform technology to obtain wavelet transient characteristics, and obtaining a target valve closing detection model by using a sound energy value, a zero-crossing rate, a Mel-frequency cepstrum coefficient, the spectral centroid, the wavelet transient characteristics and a machine learning algorithm; and detecting the to-be-detected audio based on the target valve closing detection model to obtain a detection result. And the accuracy of valve monitoring in a complex environment is improved.

The invention discloses a time-domain harmonic structure-based range expansion method for blowpipe musical instruments, and the method comprises the steps: carrying out the cross-family timbre similarity analysis: carrying out the cross-family similarity analysis of narrow-range high-pitch national blowpipe musical instruments, such as suona horn, Xun and the like, and copper pipe family and wooden pipe family musical instruments, wherein the wooden pipe family can be subdivided into an edge excitation family, a single reed family and a double reed family musical instrument, and an optimal mapping reference musical instrument of the target musical instrument is determined; adjusting the pitch and the length of each audio sample; extracting harmonic energy structures of sample audios of the reference musical instrument and other range musical instruments (medium, low and multiple bass) in the symphony configuration; calculating to obtain a multi-sample weighted frequency spectrum centroid and extracting a harmonic energy structure-based mapping rule group between the reference musical instrument and the family; extracting pure harmonics of the audio sample of the target narrow-range musical instrument according to frames, and adjusting the energy structure of each harmonic according to the mapping rule group extracted from the reference musical instrument; signals are reconstructed according to frames, and range expansion of the target narrow-range wind instrument is achieved.