639 results about "Filter bank" patented technology

The invention discloses a pan-tilt-free dynamic compensation water flow velocity measurement system used in a vibration environment. The pan-tilt-free dynamic compensation water flow velocity measurement system comprises a multi-source sensor module, a dynamic motion compensation module, a self-adaptive Kalman filter bank and a data credibility evaluation module, the multi-source sensor module is used for collecting platform attitude information, water flow original speed and platform height. The dynamic motion compensation module comprises an attitude solution sub-module and an error calculation sub-module and is used for realizing flow velocity dynamic compensation under the condition of no holder structure; the adaptive Kalman filter bank comprises an angular velocity filter and a flow velocity filter, the angular velocity filter dynamically adjusts filtering parameters according to the angular motion intensity of the platform, and the flow velocity filter dynamically switches processing strategies according to the effectiveness of a water flow signal measured by the radar; and the data credibility evaluation module is used for carrying out stability judgment based on a sliding window variance-trend joint detection algorithm. According to the invention, attitude calculation and motion compensation are fused, and accurate measurement of the water flow velocity of the holder-free platform in a vibration environment is realized.

The invention discloses a multi-mode counterfeit video detection method based on a multi-head addition cross attention mechanism, and belongs to the technical field of video counterfeit detection. The method comprises the following steps: preprocessing a video stream, decomposing a single-frame positioning face, and extracting an audio to generate a Mel spectrogram slice; the 3D convolutional network extracts video spatio-temporal features and motion differences, and the filter bank extracts audio features in combination with the residual network; audio features are mapped to a video alignment space through asymmetric projection, the video features are subjected to bidirectional interaction with an audio input multi-head addition cross attention module after being subjected to time sequence coding, and audio dominant and video dominant features are generated and are cascaded and fused with original features; and constructing cross-modal similarity loss constraint feature distribution, and fusing feature dynamic weighting and time sequence compression to output four classification probabilities of audio-visual double true, audio-visual double pseudo, video pseudo-audio true and video pseudo-audio pseudo. The multi-mode counterfeiting recognition precision is improved, and texture abnormity and audio and video mismatch features are captured.

The invention provides a dance video generation method based on multi-mode music driving and frequency domain-space double-flow decomposition, and the method comprises the steps: extracting multi-granularity music features through a composite encoder (Librosa + Jukebox), employing a beat gating attention mechanism, enabling the key actions such as dancing hand raising, kicking and the like to be strictly aligned with a music re-beat point, and enabling the synchronization error to be reduced to 118 ms through the verification of a test data set; for the problem of visual detail loss, a frequency domain-space double-flow decomposition architecture is provided, a Butterworth filter bank is used to decouple a reference image into a low-frequency energy diagram and a high-frequency residual error, and a double-flow diffusion mechanism is used to optimize a global attitude and local details respectively; a joint confidence prediction module is introduced for the generation stability in a shielding scene, and the motion trail of an abnormal joint point is dynamically corrected through a time domain sliding window weighted fusion strategy, so that a reasonable action conforming to ergonomics can still be generated under a 50% limb shielding rate.

The invention discloses a signal-to-noise ratio improvement method based on a multi-core optical fiber dual-channel distributed acoustic sensing system, and the method comprises the steps: carrying out the light splitting of light emitted by a narrow linewidth laser through a first coupler C1, enabling one path of light to generate a pulse signal through an acoustic optical modulator (AOM), enabling the other path of light to pass through a second coupler C2, and enabling the other path of light to pass through a second coupler C1; the pulse light is divided into two beams of transmission light, the two beams of transmission light are injected into a connected sensing optical fiber through two groups of erbium-doped optical fiber amplifiers (EDFA), two groups of filters, two groups of circulators and a fan-in coupler Fan-in Coupler, and the pulse light is subjected to backward Rayleigh scattering in the sensing optical fiber; based on a dual-channel DAS system, a multi-core optical fiber is used as a sensing optical fiber, three cores and four cores of the multi-core optical fiber are connected in series to form two sensing channels, seven-core space division signals are multiplexed, the equivalent self-noise pressure suppression effect of the seven-core space division signals is analyzed, after noise correlation is removed through a time domain dislocation-frequency domain phase shift method, the signal-to-noise ratio gain can be increased to 8.65 dB, and the signal-to-noise ratio gain can be increased to 8.65 dB. And the detection capability of the system on low-frequency weak signals can be improved.

The invention discloses a polar region adaptive underwater acoustic positioning correction method based on a multi-dimensional environment coupling model, belongs to the technical field of underwater acoustic positioning, is used for polar region underwater acoustic positioning, and comprises the steps of deploying a heterogeneous sensor network, constructing an environment sensing layer and eliminating a sensor drift error. Establishing a three-dimensional sound velocity model, fusing a vortex kinetic equation, a Kriging interpolation algorithm and an adaptive Runge-Kutta method to track a sound ray trajectory; designing a time-frequency double-domain joint detection algorithm, and combining a lattice IIR filter bank to realize multipath suppression; a 9-dimensional robust UKF algorithm framework is constructed, and multi-source data fusion is realized in combination with federated filtering; and terrestrial magnetism matching and a strain energy density early warning mechanism are integrated. Compared with the prior art, the method improves the positioning accuracy and stability, can more accurately predict the sound propagation path, improves the positioning precision, achieves the dynamic real-time calculation, can quickly and accurately process the continuously changing environment data, and is suitable for the complex polar region environment.

The invention discloses an equivalent self-noise pressure suppression method and device based on a multi-core optical fiber space division multiplexing signal, light emitted by a narrow linewidth laser is split through a first coupler C1, one path of light passes through an acousto-optic modulator AOM to generate a pulse signal, and then the pulse signal passes through a second coupler C2 to generate a pulse signal; the pulse light is divided into two beams of transmission light, the two beams of transmission light are injected into a connected sensing optical fiber through two groups of erbium-doped optical fiber amplifiers (EDFA), two groups of filters, two groups of circulators and a fan-in coupler Fan-in Coupler, and the pulse light is subjected to backward Rayleigh scattering in the sensing optical fiber; based on the dual-channel DAS system, the multi-core optical fiber serves as the sensing optical fiber, three cores and four cores of the multi-core optical fiber are connected in series to form two sensing channels, seven-core space division signals are multiplexed, the equivalent self-noise suppression effect of the seven-core space division signals is analyzed, and the detection capacity of the system for low-frequency weak signals can be improved.

The invention discloses a noise signal recognition method based on deep learning. The method comprises the steps of original signal collection, signal preprocessing, noise recognition feature extraction, noise signal recognition model establishment and noise signal intelligent recognition. The invention relates to the technical field of digital signal data processing, in particular to a noise signal identification method based on deep learning, and innovatively provides a self-adaptive signal mode decomposition method based on a combined judgment mechanism of energy threshold value judgment and energy range value judgment. Decoupling and reservation of different frequency bandwidth information in an original signal can be effectively completed, and the recognition accuracy of non-stationary communication noise is improved; a frequency band filter bank constructed based on a Gaussian function is introduced, so that the accuracy of communication signal identification is improved; a method for constructing a noise recognition model by fusing graph neural network structure modeling and an original type reference vector matching mechanism is innovatively proposed, and the recognition robustness of the model to non-stationary communication interference signals is improved.

The invention discloses a double-comb non-cooperative target ranging system based on single photon detection, which is characterized in that an optical pulse output by a first optical frequency comb is divided into two paths of light through a first beam splitter, the first path of light is combined with one of two paths of light divided by a second optical frequency comb through a second beam splitter, and the combined light is transmitted to a first nonlinear frequency up-conversion module; a sum frequency light pulse in the wave band of 780 nm is generated and detected by a base avalanche diode photoelectric detector, and then a time-dependent single photon counter is triggered to start counting; the second path of light is output to a beam expanding collimator through the circulator and irradiates a target to be measured through spatial propagation, the beam expanding collimator collects diffuse reflection light and returns the diffuse reflection light to a third port of the circulator along the original path, and the diffuse reflection light and the other path of light split by the second optical frequency comb through the second beam splitter are combined and transmitted to a second nonlinear frequency up-conversion module; the up-conversion light after sum frequency conversion passes through the optical filter group and then is recorded by a stop end of the single photon counting module, multi-cycle accumulative statistics is carried out by the time correlation single photon counter, and finally a photon counting waveform is reconstructed and the flight time difference of the to-be-measured target is calculated by the computer.

The invention belongs to the technical field of multimedia security, and particularly relates to a deep forgery detection method and system based on audio and video multi-mode fusion. The method comprises the following steps: extracting lip motion space-time characteristics from a video stream through dynamic ROI (Region of Interest) cutting, and processing an audio stream through fast Fourier transform and a Mel filter bank in sequence to obtain audio spectrum characteristics; reconstructing and generating a corresponding audio feature based on the spatio-temporal feature of lip motion, and fusing the audio spectrum feature and the generated audio feature by adopting a bidirectional cross-modal attention mechanism to obtain an attention fusion feature; a Mel spectrogram of the audio stream is acquired, and feature maps of different scales are extracted from the Mel spectrogram and the video stream by using a feature extraction model formed by a convolutional layer and are fused to obtain multi-scale features; after the attention fusion feature and the multi-scale fusion feature are flattened, performing weighted fusion after channel dimension splicing, and outputting a forgery probability through a multi-layer perceptron. The video detection performance is improved.

The invention discloses a ship equipment vibration test method and system based on frequency domain kurtosis mapping and non-Gaussian signal generation. The ship equipment vibration test method comprises the following steps: collecting an actual ship impact vibration signal and extracting a target power spectral density curve; the method comprises the following steps: identifying an impact energy concentration frequency band based on spectrum energy distribution, carrying out frequency band division by adopting an elliptical filter bank with a second-order section structure, and calculating an envelope kurtosis value of each frequency band to establish a kurtosis-frequency mapping model; after the initial Gaussian driving signal is generated, a non-Gaussian driving signal is generated through a phase randomization and amplitude dynamic scaling algorithm based on a mapping model; a multi-target optimization controller based on Pareto leading edge search is adopted, a power spectral density error and a kurtosis error are taken as optimization targets, synchronous optimization is carried out through fast non-dominated sorting and an adaptive hybrid variation strategy, and a driving signal is output. According to the invention, the broadband and non-Gaussian impact vibration environment of the ship equipment in the actual severe sea condition can be reproduced in a high-fidelity manner, and the authenticity of the test and the fault mode coverage capability are remarkably improved.

The invention discloses a table data frequency domain anomaly detection method based on wavelet decoupling and adaptive reconstruction, and belongs to the technical field of anomaly detection. Aiming at the problems of insufficient utilization of data relevance and neglect of frequency domain mode difference in the prior art, spectrum decoupling is realized by constructing a Beta wavelet filter bank, and detection is performed in combination with an adaptive frequency selection module and a full-band reconstruction mechanism. The method specifically comprises the following steps: (1) constructing a table graph model based on a thermonuclear function, and quantifying node relevance; (2) realizing multi-scale frequency domain decomposition by adopting a compact support Beta wavelet, and separating a normal data core mode; (3) designing an alternating encoder-attention architecture, and dynamically screening key frequency components; and (4) constructing an anomaly scoring system through the reconstruction error. Through verification of 23 cross-domain reference data sets, the AUC-ROC and AUC-PR indexes of the method are averagely improved by 8.92%, and the method is remarkably superior to the prior art. The provided Rayleigh quotient purity index provides a quantification tool for spectrum energy distribution analysis, and reveals the strong correlation between frequency selection and spectrum energy.

The invention relates to the field of rotating machinery fault diagnosis, and discloses a bearing fault diagnosis method based on a physical perception KAM network, and the method comprises the steps: carrying out the discretization of a bearing vibration signal through a Gabor filter group based on physical prior initialization, and generating modal feature lexical elements with physical frequency band meanings; and inputting the lexical elements into a PC-KAM backbone network, calculating a hidden state vector by using a state space model branch, and dynamically adjusting the position of a primary function node of a Kolmogov-Arnod network branch to realize collaborative dynamic feature extraction. In the training stage, an orthogonal subspace constraint and physical perception low-rank adaptation fine tuning mechanism is introduced. And finally, searching a historical fault case, performing multi-modal fusion on the historical fault case and the deep feature sequence, mapping a fusion representation into a soft prompt, and inputting the soft prompt into a large language model to generate a diagnosis report. According to the method, the problems of poor physical interpretability of characteristics and few-sample diagnosis under variable working conditions are effectively solved, and the generalization and decision-making ability of a diagnosis system are improved.

The invention relates to the technical field of voiceprint recognition, and discloses an air-ground conversation speaker recognition method for short voice and complex noise, which dynamically suppresses background noise through a self-adaptive filter in combination with a minimum mean square algorithm, and improves the voice signal quality. A multi-scale feature fusion technology is adopted, short-time Fourier transform, filter bank energy features and multi-scale convolution are combined, a random time and frequency shielding data enhancement strategy is introduced, and the short speech feature expression ability is enhanced; bi-GRU is integrated in the ECAPA-TDNN model, and context information of multiple rounds of conversations is captured through context time sequence modeling, so that the recognition stability in a dynamic scene is improved; and an AAM-Softmax loss function is adopted to strengthen the distinguishing capability of the model, and finally identity judgment is realized through cosine similarity matching. According to the method, the voiceprint recognition precision and robustness under complex noise and short voice conditions are remarkably improved, and the safety risk caused by identity misjudgment in aviation communication is effectively reduced.

The invention discloses an acousto-optic combined detection and fusion positioning method and device for underground gas leakage, and belongs to the technical field of industrial safety monitoring. In order to solve the problems of high false alarm rate, insufficient positioning precision, poor dynamic environment adaptability and the like of a single sensor in the prior art, the invention provides a multi-modal signal space-time collaborative fusion technical scheme which specifically comprises an acousto-optic heterogeneous sensor, an integrated MEMS ultrasonic array and a tunable laser spectrum module. Microsecond-level synchronous acquisition of acousto-optic signals is realized through a GPS taming clock, and a cross-correlation time delay compensation model is designed to eliminate space-time mismatch errors. And designing a three-stage dynamic fusion framework: improving LMS adaptive filtering to suppress industrial noise, establishing a light intensity attenuation model in combination with dust environment concentration, carrying out multi-modal feature extraction of acousto-optic signals, carrying out signal preliminary fusion, and judging a gas leakage condition. And constructing an acousto-optic Bayesian joint probability model, and calculating the leakage posterior probability. And optimizing positioning parameters based on a double-Kalman filter group, predicting sound wave time delay by forward filtering, and correcting a light diffusion coefficient by backward filtering to realize three-dimensional coordinate solution. Temperature, humidity and dust concentration parameters are fused in real time, a sound velocity calculation model and a light intensity attenuation factor are corrected, and the robustness of complex working conditions is improved. According to the invention, the advantages of acousto-optic signals are combined, errors of various complex environmental factors are further fused and compensated, the overall error is greatly reduced, and the method has the characteristic of dynamic adaptation, and is very suitable for gas leakage detection and positioning in complex environments such as underground environments and the like.

The invention relates to a time domain broadband adaptive beam forming method and system based on Chebyshev weighting, and belongs to the field of array signal processing. Aiming at the technical problems of poor real-time performance and high sidelobe of the existing broadband beam forming, the method comprises the following steps of: dividing a received broadband signal into a plurality of sub-bands, obtaining an optimal weight vector for each sub-band by adopting a minimum linear constraint iterative algorithm after down-conversion, constructing an expected response of an FIR filter according to each weight, obtaining a filter coefficient through L2 norm convex optimization, and finally obtaining a high sidelobe beam according to the expected response of the FIR filter. And performing time domain weighting on each array element signal by using a filter bank, and outputting a directional beam. According to the method, the calculated amount is low, a main lobe and interference null can be formed in real time in a broadband, and a certain side lobe height is kept.

The invention belongs to the technical field of speech recognition and acoustic modeling, and particularly relates to a high-fidelity end-to-end acoustic modeling method combining speech intelligibility reconstruction and autoregressive feedback optimization, which comprises the following steps of: firstly, establishing an intelligibility loss function based on a human speech intelligibility model; subjective definition features of the voice are reconstructed through a frequency spectrum reconstruction network, intelligibility related features are extracted and embedded in combination with a perceptual weighted filter bank, and the intelligibility related features and Mel frequency spectrum features are input into an acoustic encoder in parallel; based on this, by introducing a voice intelligibility reconstruction and autoregressive feedback mechanism, while the conciseness of an end-to-end voice modeling framework is maintained, systematic improvement of voice sharpness, naturalness and stability is realized, and the method can be widely applied to scenes such as intelligent voice assistants, voice transcription, virtual anchors, voice restoration, cross-language voice generation and the like. And the method has extremely high practical application value and popularization prospect.

The invention discloses an equivalent self-noise pressure suppression method based on a dual-channel distributed acoustic sensing system, and the method comprises the steps: carrying out the light splitting of light emitted by a narrow-linewidth laser through a first coupler C1, enabling one path of light to generate a pulse signal through an acoustic optical modulator (AOM), enabling the other path of light to pass through a second coupler C2, and enabling the other path of light to pass through a second coupler C1; the pulse light is divided into two beams of transmission light, the two beams of transmission light are injected into a connected sensing optical fiber through two groups of erbium-doped optical fiber amplifiers (EDFA), two groups of filters, two groups of circulators and a fan-in coupler Fan-in Coupler, and the pulse light is subjected to backward Rayleigh scattering in the sensing optical fiber; according to the invention, based on a dual-channel DAS system, two fiber cores of the multi-core optical fiber are used as sensing optical fibers, and the space division signals of the two cores are multiplexed, so that the performance that the equivalent self-noise pressure drop is 2.54 dB in the frequency band of 100-1000Hz is realized under the condition that the noise correlation coefficient of the dual-core signals is 0.11, and the detection capability of the system on low-frequency weak signals can be improved.

The invention discloses a wind driven generator large-torque transmission chain resistance increasing control method and system, and relates to the technical field of wind driven generator resistance increasing control. The method comprises the steps that operation parameters of a wind driven generator are collected in real time and preprocessed; calculating a rotation fluctuation frequency set and a corresponding amplitude spectrum of the current time window, and tracking and locking the dominant vibration frequency; constructing a fan state space prediction model containing the torsional vibration state of the transmission chain, outputting the predicted variable quantity of the torque of the transmission chain in the next control period, constructing a filter bank based on the dominant vibration frequency, performing filtering compensation on the predicted variable quantity, and generating an electromagnetic torque given value; and performing phase lead compensation on the electromagnetic torque given value, generating a control instruction, sending the control instruction to the wind power converter, and controlling the wind power generator to output the corresponding electromagnetic torque by using the wind power converter. Through systematic multi-link cooperative processing, the vibration suppression effect and the operation stability of the transmission chain can be comprehensively improved.

The embodiment of the invention discloses a broadband signal simulation method and system influenced by Doppler frequency in satellite communication. The method comprises the following steps: sampling a received original broadband signal according to an original sampling rate, and obtaining a signal value of a sampling point in real time; discretizing the sampling time of the original broadband signal at the original sampling rate and after the original broadband signal is influenced by the Doppler effect to obtain a first discrete time value and a second discrete time value; storing the signal value and the first time value in a plurality of filter banks in parallel; based on the second time value and the first time value, triggering the filter bank to perform interpolation operation on the stored signal value to obtain a signal interpolation; and performing frequency modulation on the original broadband signal based on the carrier frequency, the relative speed and the output signal interpolation to form a target signal. According to the method, the signal conversion process can be simulated through the simulation technology, the low-cost performance test requirement of the satellite communication system is met, and meanwhile processing time and resource consumption are reduced according to the multiple filter banks.

The invention discloses a voice interactive chart dynamic generation method and system based on a large language model, and relates to the technical field of chart generation, and the method comprises the steps: collecting a user voice signal, carrying out the preliminary framing, calculating the short-time energy, and analyzing a mute segment set signal; frequency domain signals are extracted from the set signals, gain is calculated through a filter, noise reduction is carried out on the frequency domain signals, the zero-crossing rate and noise reduction short-time energy are calculated, effective voice frames are screened, peak detection is carried out, the average voice speed is calculated, voice speed normalization is carried out, and signal values of sampling points in the frames are extracted. According to the method, the noise power spectrum density is extracted through FFT on the mute section, a frequency domain model of background noise is effectively established, a follow-up filter can accurately act on an actually existing frequency band interference area, weakening of the voice main signal spectrum is avoided, the Mel filter bank is accessed after the power spectrum is calculated through the FFT after frame windowing, and the noise power spectrum density is calculated through the FFT after frame windowing. Energy can be redistributed on the logarithmic Mel scale according to human ear perception characteristics.

The invention relates to the technical field of microwave communication, and discloses a microwave communication radio frequency front-end transceiver chip and a dynamic bandwidth allocation method and system. The system comprises a radio frequency front-end transceiver chip; the control module is electrically connected with the radio frequency front-end transceiver chip and is configured to receive the digital signal sent by the analog-to-digital conversion module so as to extract signal characteristic parameters, and the signal characteristic parameters comprise a signal-to-noise ratio, a multipath fading coefficient and a target scene communication protocol parameter; determining an initial bandwidth allocation strategy and a power control threshold according to the target scene communication protocol parameters; based on the signal-to-noise ratio and the multipath fading coefficient, the working frequency band and the passband width of the multi-frequency-band filter bank are dynamically adjusted through an adaptive particle swarm optimization algorithm, and a gain control signal of the power amplification module is generated, so that the transmitting power is minimized on the premise of meeting the communication distance requirement, and dynamic bandwidth allocation is completed. And the performance and the adaptability of microwave communication in a complex scene are remarkably improved.

Apparatus for retrieving an acoustic room information from a database, comprising: a test fingerprint processor (10) for deriving a test fingerprint from a test acoustic room information; and a database processor (20) for retrieving a reference acoustic room information having a reference fingerprint matching with the test fingerprint, wherein the test fingerprint processor (10) comprises; a filter bank configured for decomposing the test acoustic room information into a plurality of acoustic room information sub-bands; and a parameter calculator (12) configured for calculating a sub-band parameter for each acoustic room information sub-band of the plurality of acoustic room information sub-bands, the sub-band parameter indicating an energy change over time in the acoustic room information sub-band, wherein the test fingerprint is based on the sub-band parameters for the acoustic room information sub-bands.

The invention relates to the technical field of audio processing, and discloses an intelligent control system and method for a wireless microphone, and the method comprises the steps: collecting multi-point acoustic data through a distributed microphone network, and constructing a spatial acoustic topology model; based on a topological analysis result, constructing an adaptive notch filter bank to realize howling suppression; environment changes are monitored in real time through a distributed acoustic sensor network, and suppression parameters are dynamically optimized; integrating a lightweight convolutional neural network on microphone equipment to identify acoustic events; low-delay acoustic event processing and howling prevention are realized by adopting a layered edge computing architecture; according to the invention, the technical problems of unstable howling suppression, weak environmental adaptability and lack of spatial perception in a complex environment are effectively solved.

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 discloses a micro eye jump detection method and device, near-eye display equipment and a computer readable storage medium. The micro eye jump detection method comprises the following steps: acquiring original data, wherein the original data comprises a time-varying coordinate sequence of an eyeball fixation point; processing the original data based on the filter bank to judge whether micro-eye jump motion exists in the coordinate sequence or not; and controlling the near-to-eye display equipment to execute corresponding operation according to the judgment result. Therefore, noise interference can be effectively avoided, the detection accuracy is improved, the calculation overhead can be reduced, and real-time detection is realized.

The invention provides a ship radiation noise identification method and system based on a double-low-rank adjustment network. A ship radiation noise identification method based on a double-low-rank adjustment network comprises the following steps: performing Mel filter bank transformation on a ship radiation noise signal to obtain a Mel time-frequency spectrogram of the noise signal; inputting the Mel time-frequency spectrogram into a trained noise recognition model, and outputting a target category of the ship radiation noise; the noise recognition model is generated based on a pre-trained reference model containing a plurality of residual blocks, low-rank adaptation is performed on a weight space through a weight low-rank adaptive module, and low-rank adaptation is performed on a feature space through a feature low-rank adaptive module. The method has the advantages that the classification performance of ShipsEar and DeepShip data sets can be remarkably improved under the condition that only 0.55% of parameters are increased, and the method has low overhead and high effectiveness.

The present invention provides a method and system for detecting the switching state of the closing resistor of an AC filter group circuit breaker, which belongs to the technical field of high-voltage AC circuit breakers. The method comprises: loading the closing current recording data of the filter group circuit breaker switching process; performing wavelet decomposition on the closing current recording data of the circuit breaker, obtaining the singular points in the closing current recording data of each layer of the circuit breaker as the starting points of each transient process during the closing resistor switching process of the circuit breaker; performing TLS-ESPRIT decomposition on the recording data of each transient process to obtain the amplitude and attenuation factor of the DC mode, and establishing a corresponding relationship between the amplitude, attenuation factor and the actual resistance value of the closing resistor; fitting two amplitude-resistance curves and two attenuation factor-resistance curves according to the corresponding relationship, as well as the curve function and inverse function of each curve; substituting the amplitude and attenuation factor into the inverse function of the corresponding DC component to calculate the resistance value, and further calculating the number of flashover resistors based on the average resistance value, so as to understand whether a flashover fault occurs.

The invention relates to an underwater acoustic communication and positioning integrated signal design and processing method based on FDMA signal modulation, a medium and a device, and belongs to the field of underwater acoustic communication / positioning. And respectively modulating the integrated baseband signals to a plurality of pre-divided carriers with different center frequencies. Each carrier corresponds to an independent transmitter signal, and the center frequency interval of different transmitter signals is greater than the signal bandwidth. And the modulated signal is transmitted to an underwater environment through the underwater acoustic transducer. In the signal processing process, a band-pass filter bank is used for carrying out frequency separation on received signals, each filter corresponds to the center frequency of a transmitting signal, the separated signals are demodulated and decoded to recover original communication data, and the distance between a transmitter and a receiver is calculated by measuring time delay. And simultaneous communication interaction and high-precision positioning of multiple underwater users are realized.

The invention discloses a sound event detection method and system based on class distribution and time sequence context collaborative prompt, and the method comprises the steps: converting an original audio signal into a signal frame sequence, extracting the features of a Mel filter bank through a pre-training model branch, and extracting the features of a Mel spectrum through a downstream model branch; a combination of a local time sequence prompt module and a global distribution prompt module is introduced into each layer of the pre-training model, and the pre-training model outputs audio sequence features and the global distribution prompt module; performing feature fusion on the audio sequence and the output of the downstream model, calculating the frame level prediction probability of the downstream model, and realizing an audio positioning task; generating a sentence level prediction probability from the frame level prediction probability of the downstream model; for the pre-training model, processing the global distribution prompt module to obtain a sentence level prediction probability of the pre-training model; and finally, fusing the two sentence level prediction probabilities to obtain a classification result of the sound event. According to the method, the audio classification and positioning performance of sound event detection can be remarkably improved.

The invention provides an inter-channel crosstalk suppression method in a multi-channel optical module, which aims at the problems of inter-channel interference and signal overlapping caused by wavelength shift in the multi-channel optical module, calculates the wavelength shift by adopting a proportional integral differential control method through collecting the reading of a temperature sensor and the output wavelength data of a laser device in real time, and finally suppresses the crosstalk in the multi-channel optical module. The threshold value is dynamically adjusted in combination with the environment temperature change, and the interference risk index is quantified; the wavelength drift trend is extracted through time sequence analysis, the channel interval adjustment logic is optimized, the filter bank is dynamically configured to improve the signal purity, meanwhile, the wavelength parameter is corrected through the offset compensation value, and the stability is ensured. If the transmission quality requirement is met, configuration parameters are locked, a correction model is matched and optimized through historical data, potential risks are predicted in combination with state estimation filtering, and the logic adaptability is enhanced and adjusted. According to the invention, efficient correction of wavelength drift and interference minimization are finally realized, and the transmission quality and operation stability of the multi-channel optical module are improved.