24 results about "Signal statistics" patented technology

Technology is disclosed for an optical receiver. The optical receiver may include an optical and digital signal processing (ODSP) device. The ODSP device may include a photodiode operable to convert an optical signal to an electrical signal; a digital signal processor operable to generate a digital signal based on the electrical signal; a quantization component operable to generate a quantized output signal based on the digital signal; and a processing device. The processing device may be operable to measure one or more signal statistics of the digital signal; and identify one or more quantization thresholds, wherein the one or more quantization thresholds are computed based on the one or more signal statistics.

The present application relates to the technical field of energy storage battery, in particular to a forward-looking energy storage frequency modulation power distribution method and system; the method has a double-time scale structure of minute-level strategy adjustment + second-level real-time control. On the minute-level time scale, an AGC signal statistical feature extraction and prediction framework is constructed to predict the AGC signal characteristics and dynamically adjust and optimize the weight; on the second-level time scale, multi-objective optimization is performed every 2 seconds to distribute the charging and discharging power among the battery units, minimizing the AGC tracking error, battery degradation and SOC deviation. The scheme can significantly reduce the battery degradation cost while maintaining good frequency modulation performance and SOC consistency; by capturing the time statistical characteristics of the AGC signal, the potential impact of the AGC signal on the battery degradation is quantified, that is, the influence of the change trend of the AGC signal on the battery energy storage system is considered, providing forward-looking support for the control strategy of power distribution.

The embodiment of the present disclosure provides a loudspeaker audio classification method under a plurality of heterogeneous devices, and belongs to the technical field of data identification, and specifically comprises the following steps: step 1, calling a built-in acceleration sensor of a mobile phone to collect an initial vibration signal played by a same on-board loudspeaker; step 2, performing a pretreatment operation on the initial vibration signal; step 3, removing interference signals in the pretreated initial vibration signal to obtain a target vibration signal; step 4, screening signal statistical features highly related to audio content and independent of devices; and step 5, classifying the audio content of the target vibration signal by using an adversarial neural network according to the signal statistical features. Through the scheme of the present disclosure, the efficiency and accuracy of speech recognition are improved.

The application discloses a valve tester verification method and system of reverse resistance type IGCT driving function, and relates to the technical field of direct current transmission.The valve control host detects the driving up-sent IGCT forward voltage signal, and the delay input / exit test mode is put into operation.The valve tester is divided into multiple rounds of statistical stages and trigger stages, and the forward voltage signal statistics are realized, so that multiple test function verifications are realized.The valve tester verification method and system of reverse resistance type IGCT driving function are aimed at the characteristics that the IGCT driving energy taking time is long and cannot be automatically turned off, realize the effect that the valve tester identifies test items and matches control time sequences under the condition that the valve tester does not directly communicate with the valve control host or the IGCT driving, and satisfies the test requirements of comprehensive testing of the reverse resistance type IGCT driving performance.

Techniques for an optical receiver are disclosed. The optical receiver may include an optical digital signal processing (ODSP) device. The ODSP device may include: a photodiode operable to convert an optical signal into an electrical signal; a digital signal processor operable to generate a digital signal based on the electrical signal; a quantization component operable to generate a quantized output signal based on the digital signal; and a processing device. The processing device is operable to: measure one or more signal statistics of the digital signal; and identifying one or more quantization thresholds, wherein the one or more quantization thresholds are calculated based on the one or more signal statistics.

The present application belongs to the field of aviation manufacturing and intelligent processing technology, and discloses a laminated interface online identification method based on an adaptive statistical discrimination mechanism. The method preprocesses the machining state signal, judges whether to enter the stable drilling machining stage based on the change characteristics of the machining state signal in the sliding time window, and calculates the dynamic mean value of the machining state signal in the stable drilling machining stage. The bilateral cumulative sum statistics are constructed based on the dynamic mean value, and the interface is judged in combination with the first and second judgment thresholds which are adaptively adjusted based on the statistical characteristics of the historical signal. When the preset interface judgment condition is met, it is judged that the machining state at the current drilling position has a sudden change, and the interface position is identified. The method does not need to establish a complex signal model and parameter calibration, can adaptively compensate the slow trend in the machining process, and is suitable for interface online identification under complex working conditions of multi-material laminated structure.

The application relates to the technical field of visual detection, and discloses a drowning prevention early warning method and system based on visual detection, which comprises the following steps: acquiring a water surface video and generating an optical flow distribution map; separating a foreground motion region based on optical flow statistical characteristics; extracting reflection noise features for denoising treatment to obtain pure motion data flow; performing time-frequency and extreme value analysis on the pure motion data flow to determine preliminary signs of struggling behavior; performing displacement analysis on multiple frames of images to obtain abnormal behavior tracking results; accumulatively calculating a comprehensive score of a drowning event, and generating an early warning signal when the score exceeds a threshold value; and counting the number of receiving terminals, and triggering an event processing procedure when the number of receiving terminals meets a threshold value. The method can effectively suppress water surface reflection interference and realize high-reliability drowning prevention early warning.

The present application relates to the technical field of physical testing, in particular to an ultrasonic flaw detection method, system and equipment for isostatic pressing graphite finished products, the method comprising: determining local fluctuation synchronization features of a target position based on ultrasonic time domain signals of the isostatic pressing graphite finished product at the target position and its spatiotemporal neighborhood positions; determining multi-scale scattering correlation features of the target position by using mutual information between adjacent scale component signal features of the ultrasonic time domain signals; determining scattering events based on local signal-to-noise ratios of the ultrasonic time domain signals, determining event state transition probability matrices based on scattering event features, and obtaining sequence signal statistical evolution features of the target position; determining a comprehensive structure continuity parameter of the target position by combining the above multiple features of the target position, and obtaining a flaw detection result by adjusting separation constraint criteria of ICA based on the comprehensive structure continuity parameter. Through the technical scheme of the present application, substantial improvement in accuracy and reliability of graphite material defect detection is achieved.

This invention discloses a method and system for predicting industrial equipment faults based on deep learning, belonging to the field of fault diagnosis and prediction technology. The method includes: constructing an equipment topology graph model, abstracting sensor data channels as graph nodes, and constructing an adjacency matrix based on physical mechanical connections and signal statistical correlations to characterize stress transmission and vibration coupling paths; extracting spatial feature vectors through a multi-layer graph convolutional neural network in a parallel architecture, and simultaneously extracting temporal feature vectors through a multi-head self-attention mechanism; concatenating the spatiotemporal features and extracting equipment health feature vectors through a nonlinear mapping module containing parameterized linear rectifier units; outputting the posterior probability of the fault and triggering an early warning. The system includes a topology graph model construction module, a parallel feature extraction module, a feature fusion and mapping module, and a fault prediction output module. This invention can achieve highly sensitive identification and early warning of early faults in industrial equipment.

The invention discloses a partial discharge identification method and device based on multi-dimensional waveform feature extraction, and belongs to the technical field of state monitoring and intelligent diagnosis of high-voltage electrical equipment, and the method comprises the steps: obtaining the voltage and current waveform data of a transformer under the impact voltage, and carrying out the preprocessing of the voltage and current waveform data, voltage and current waveform preprocessing data are obtained; original signal statistical characteristics, partial discharge event characteristics and voltage structure characteristics are extracted; standardized features are obtained through standardized processing; constructing a neural network model based on the standardized features; training the neural network model by using a labeled sample set to obtain a trained neural network model; and inputting a signal acquired in real time into the trained neural network module for calculation so as to realize identification of the partial discharge signal. According to the invention, the features are trained and classified by using the multilayer perceptron neural network, and automatic identification of different discharge stages is realized.

The invention provides a dose determination method and device based on a time sequence signal, equipment and a storage medium, and relates to the field of physics and computers. The method comprises the following steps: acquiring a plurality of sample CT images and determining dose distribution information and nuclide distribution information of each sample CT image; for each sample CT image, generating a PET accumulated signal image corresponding to each signal statistics window based on nuclide distribution information; training the first initial model by taking the sample CT image and the PET accumulated signal image corresponding to the sample CT image as a first training data set to obtain a first target model; training a second initial model by taking the dose distribution information, an output result of the first target model and part of training data in the first training data set as a second training set to obtain a second target model; and sequentially inputting the target CT image and the PET accumulated signal image of the target time window into the first target model and the second target model to obtain dose information output by the second target model.

The invention relates to the field of pipeline leakage detection, and discloses a digital leakage listening detection system based on stepless frequency and multi-band bandwidth adjustment, the system comprises a signal acquisition module, a user adjustment module, a main control module, a display module, an audio output module and a power supply module, and the system constructs a real-time simulation auditory sense and asynchronous digital vision dual-channel architecture; the core of the method is that a main control module calculates a leakage signal steady-state index by fusing an energy stability index representing signal energy time stability and a statistical randomness index representing signal statistical randomness in an asynchronous digital visual path; the index provides an objective and quantitative visual navigation basis for operators, and continuous leakage signals and transient and periodic interference can be effectively distinguished. According to the invention, auditory experience and objective data analysis of operators are considered, and the accuracy, reliability and efficiency of leakage detection are remarkably improved.

This application discloses a method, apparatus, medium, and product for estimating network packet loss rate, relating to the field of transmission and bearer technology. The method includes: acquiring signaling statistics data collected at a preset collection point in a first link, including the total number of request packets in a first direction and a second direction of the first link, and the number of retransmissions of request packets in the corresponding directions; wherein the first link includes a first end and a second end; calculating a first overall packet loss rate in the first direction and a second overall packet loss rate in the second direction of the first link based on the signaling statistics data; determining the segmented packet loss rate of segmented links of the first link based on the first overall packet loss rate and the second overall packet loss rate, wherein the segmented links include a first segmented link between the first end and the preset collection point; and a second segmented link between the second end and the preset collection point. The solution of this application solves the problem that existing packet loss rate assessment schemes cannot accurately and quantitatively estimate packet loss rates.

This invention relates to the field of mechanical equipment condition monitoring technology, and discloses a vibration signal denoising method and system based on a self-supervised convolutional neural network. The method involves preprocessing the original vibration signal during the operation of the mechanical equipment to train a constructed one-dimensional convolutional-deconvolutional neural network model. Instead of relying on manually labeled clean samples, it utilizes signal statistical characteristics such as kurtosis, envelope spectrum sparsity, and root mean square amplitude to construct a self-supervised loss function. This allows the model to automatically learn to distinguish between impact components and noise, guiding the one-dimensional convolutional-deconvolutional neural network model to automatically separate impact components from noise. The convolutional-deconvolutional structure enables temporal feature extraction and signal reconstruction, thereby achieving signal denoising while maintaining consistency with the original amplitude and effectively enhancing impact characteristics.

A circuit includes a first digital pre-distortion (DPD) corrector and a second DPD corrector. The first DPD corrector has an input, and an output. The second DPD corrector has an input coupled to the input of the first DPD corrector, and an output. A signal combiner has a first input coupled to the output of the first DPD corrector, a second input coupled to the output of the second DPD corrector, and an output. The second DPD corrector is configured to provide a signal at the output of the second DPD corrector based on a signal at the input of the second DPD corrector and one or more signal statistics related to the signal at the input of the second DPD corrector.

The invention relates to the technical field of physical testing, in particular to an ultrasonic flaw detection method, system and equipment for an isostatic pressing graphite finished product, and the method comprises the following steps: based on ultrasonic time domain signals of the isostatic pressing graphite finished product at a target position and a space-time neighborhood position thereof, determining local fluctuation synchronism characteristics of the target position; determining a multi-scale scattering correlation feature of the target position by using mutual information between adjacent scale component signal features of the ultrasonic time domain signal; determining a scattering event based on a local signal-to-noise ratio of the ultrasonic time domain signal, determining an event state transition probability matrix based on scattering event characteristics, and obtaining a sequence signal statistical evolution characteristic of the target position; the comprehensive structure continuity parameters of the target position are determined by combining the multiple features of the target position, and the flaw detection result is obtained by adjusting the separation constraint criterion of the ICA based on the comprehensive structure continuity parameters. By means of the technical scheme, the accuracy and reliability of graphite material defect detection are substantially improved.

This invention relates to the field of data processing technology and discloses a multi-granularity respiratory signal statistical inference and sleep staging decision-making system. The system comprises a signal acquisition and buffering module that segments a continuous respiratory acoustic signal sequence into continuous analysis windows; a multi-dimensional feature extraction module that extracts multi-dimensional observation feature vectors containing respiratory rate variation features and cross-window respiratory trend features; a dynamic transition probability generation module that acquires global sleep onset time parameters and continuous dwell time parameters, performs numerical updates on the initial state transition probability matrix, and generates a dynamic state transition probability matrix; and inputs the multi-dimensional observation feature vectors and the dynamic state transition probability matrix into a Markov model to calculate the posterior probability and output the classification result. This invention overcomes the aftereffect defect of static Markov models by reconstructing the state transition matrix through spatiotemporal dual variables, achieving accurate anti-disturbance and state locking against transient noise disturbances.

The application provides a data fusion processing method for frequency use behavior description based on adaptive reachable density, signal frequency parameters, signal statistical spectrum, signal communication parameters and signal direction degree are calculated through intermediate frequency spectrum, IQ and measurement data in spectrum monitoring; parameter measurement results of single sensing position and single measurement are formed by reorganizing signal frequency parameters, signal statistical spectrum, signal communication parameters and signal direction degree based on sensing position, collection time and collection frequency; signal parameters of single sensing position and single measurement are extracted from the parameter measurement results, signal parameters are classified and normalized through adaptive reachable density clustering, and normalized results of signal parameters of single sensing position and at least twice measurement are formed; signal parameter fusion is performed on each electromagnetic target based on the normalized results of signal parameters of single sensing position and at least twice measurement and corresponding electromagnetic targets. The problem that monitoring results of multiple measurements and multiple sensing positions cannot be fused is solved.

The invention discloses an audio reverberation space clue parameter estimation method and device and a storage medium, and relates to the technical field of audio processing. The method comprises the following steps: acquiring a target wet sound signal, and carrying out dry and wet sound separation on the target wet sound signal to obtain a dry sound signal and a pure reverberation signal; obtaining parameter estimation rules corresponding to a plurality of target audio space cue parameters used for representing reverberation space characteristics; the parameter estimation rule is pre-constructed based on the correlation between the signal statistical characteristic of the audio signal and the audio space clue parameter; and according to the parameter estimation rule, analyzing the target wet sound signal, the dry sound signal and the pure reverberation signal to obtain a target audio space clue parameter of the target wet sound signal. Automatic audio reverberation space clue parameter estimation can be realized, the audio space clue parameter estimation efficiency and accuracy are improved, and the method is suitable for linear and nonlinear reverberation effects.

The invention provides a multilayer adaptive rate distortion optimization coding method and device, and a storage medium. In one example, the method comprises the steps of performing semantic recognition on a current video frame, and determining a semantic importance score of each target; determining a frame-level semantic importance score according to the semantic importance score of each target, and determining a frame-level scene dynamic score according to the scene perception information of the current video frame; determining a frame-level comprehensive importance score; determining a reference QP of the current video frame according to the frame-level comprehensive importance score; determining the QP offset of each image block according to the signal statistical characteristics of each image block in the current video frame, and determining the QP of each image block according to the reference QP and the QP offset; and coding the current video frame according to the QP of each image block. The method can optimize the balance between the compression efficiency and the video quality.

The invention provides an intelligent pet necklace based on unconstrained inertial sensing and a physiological signal processing method thereof. The intelligent pet necklace is worn on the neck of a pet in an unconstrained mode, a gap change of 0-30 mm is allowed to exist between the necklace and the skin, and the necklace is allowed to rotate freely relative to the body. An inertial measurement unit is arranged in the necklace to collect multi-axis acceleration data, and the multi-axis acceleration data is converted into scalar signals irrelevant to the wearing direction through spatial irrelevant fusion; self-adaptive frequency domain adjustment is carried out according to the physiological features of the target pet, and motion noise and breathing interference are inhibited; further separating cardiovascular micro-vibration components from the mixed signals based on signal statistical independence characteristics through a blind source separation algorithm; and finally, cardiovascular activity parameters including heart rate, heart rate variability, heart beat interval sequence and the like are extracted. According to the method, high-robustness cardiovascular micro-vibration signal extraction and health monitoring are realized only by depending on a general inertial sensor and an advanced signal processing algorithm under the conditions of random wearing posture and uncertain direction.

The invention discloses a multi-modal feature fusion and recognition method fusing knowledge graph constraints, belongs to the technical field of computer data processing and multi-modal intelligent recognition, can be used for mobile terminal remote depression auxiliary screening, online psychological counseling and intelligent follow-up assessment, and adopts the scheme that audio, video and text features are extracted, and the observation quality is assessed; calculating fingerprints of an end-side processing link and a transmission transfer link based on a signal statistical form and link statistics, and evaluating and adaptively calibrating atlas constraint intensity edge by edge; and quantizing edge-level conflicts in fusion reasoning, attributing and distinguishing observation degradation and relation failure, respectively adjusting modal contribution degree coefficients or edge constraints, and forming edge-level ledgers to support sorting positioning and rollback recovery. According to the technology provided by the invention, the problems of depression risk misjudgment, difficulty in attribution of a conflict source, difficulty in positioning a failure relationship and the like caused by link-induced false evidence are solved, so that the stability and maintainability of an identification result and a risk trend under cross-equipment, cross-environment and weak network conditions are improved.

The invention discloses a borborygmus signal classification method based on an adaptive threshold value and a priority strategy, and the method comprises the steps: 1, obtaining a borborygmus signal data set, and carrying out the short-time Fourier transform of the data set, and obtaining a sound intensity sequence; 2, calculating signal statistical characteristics (a mean value and a standard deviation), and calculating an adaptive detection threshold value based on the signal statistical characteristics; 3, sequentially detecting a single burst (SB) event, identifying a multiple burst (MB) event sequence, and detecting a continuous random sound (CRS) event; 4, performing merging processing on the event intervals of the same type, and performing priority swallowing processing on the event intervals of different types; and 5, outputting a final borborygmus event classification detection result. The self-adaptive threshold value method is adopted, parameters do not need to be set manually, automatic classification detection of three types of borborygmus events of SB, MB and CRS is achieved, the problem of separation and recognition when different types of events are adjacent is systematically solved through the priority strategy, and the method has the advantages of being high in automation degree and high in classification accuracy.

The application discloses a task difficulty adaptive industrial robot health monitoring method and system, a storage medium and a computer device, and the method comprises the following steps: collecting joint bearing rotating speed and vibration signals of a target industrial robot in real time, calculating physical deviation and distribution difference between a current working condition and each preset typical working condition according to the joint bearing rotating speed and vibration signals, and determining a task level of health monitoring tasks according to the physical deviation and the distribution difference; determining a target reference working condition according to the task level, calculating physical deviation between the current working condition and the target reference working condition, generating physical deviation features according to the physical deviation, generating signal statistical features according to the vibration signals, and splicing the physical deviation features and the signal statistical features to obtain a combined feature vector; calculating scores of each candidate health monitoring model according to the combined feature vector; if the task is a simple task, performing health state monitoring according to the candidate health monitoring model with the highest score; and if the task is a complex task, performing online migration training on the candidate health monitoring model corresponding to the target reference working condition, and performing health state monitoring after the training.