8 results about "Separation algorithm" patented technology

The invention provides an automatic processing method for real-time observation data of an ocean station, and belongs to the technical field of ocean observation data processing. A wavelet packet decomposition multi-scale noise separation algorithm is established to distinguish environmental noise and real signals, dual-sensor redundancy configuration is combined with Bayesian inference to identify sensor drift, and a calibration coefficient is updated in real time through a recursive least square method. A one-dimensional time sequence is mapped to a high-dimensional phase space by utilizing a phase space reconstruction algorithm to realize high-precision prediction of a chaotic signal, a hierarchical data storage architecture is established, and a data migration strategy is iteratively optimized through a hierarchical correlation degree function; the technical problem that time synchronization signals are difficult to reconstruct accurately according to asynchronous sampling data of multiple sensors of an ocean station is solved.

The present invention relates to the fields of PID control and Kalman filtering, and disclosed is an improved PID control method based on Kalman filtering. The present invention comprises: in the process of acquiring a target value of an expected signal, taking the difference between the target value and a current feedback quantity of a controlled object to obtain an error value, acquiring an output of PID from the error value, applying the output of PID to the controlled object to obtain a noisy output signal, then, after Kalman filtering, feeding the output back to PID, continuously iterating to make the feedback quantity approach the target value, and finally obtaining a stable output signal. Variable-speed integration and integral separation algorithms are added to an integral term of PID to change an integration speed on the basis of a system deviation, and an incomplete differential algorithm and a low-pass filter are added to a differential term of PID to prevent high-frequency interference. According to the present invention, the optimal estimation of a system state is obtained by means of prediction of a current state and updating of observation data, thereby reducing the influence of measurement errors and noise, and improving the performance and stability of a PID controller.

This invention provides a multimodal fusion-based method for correlation analysis of deep-sea environmental data outside the Yangtze River Estuary, belonging to the field of deep-sea environmental analysis technology. This invention achieves accurate data registration by establishing a spatiotemporal standardization model for multi-source data and employing an adaptive spatiotemporal kriging interpolation algorithm. It utilizes a sparse coding marine signal separation algorithm to construct an overcomplete dictionary to separate mixed signals and extract pure features. Based on an attention mechanism, a feature extraction network is constructed to automatically learn deep feature representations of physical, chemical, and biological parameters. A matrix rank deficiency detection algorithm is used to automatically supplement feature compensation vectors to ensure feature integrity. A cross-scale attention mechanism based on wavelet transform is constructed to fuse multi-scale features. A Shapley value interpretability evaluation system is established to quantify the importance of correlated features. This invention solves the technical problem of inaccurate correlated feature extraction during the fusion of multi-source heterogeneous marine environmental data.

The invention relates to the technical field of digital signal processing, in particular to a leak detector anti-interference method based on an estimated background value, which comprises the following steps: acquiring a background separation algorithm of a leak detection signal during leak detection, performing derivation and filtering on the leak detection signal twice, and performing integration twice again to extract a background natural attenuation part signal; a linearization algorithm of the exponential signal: taking a natural logarithm ln from the exponential signal to linearize the exponential signal, and estimating a background value derivative algorithm; according to the method, the future derivative value is estimated through less calculation, the future derivative is estimated, the advanced derivative can be obtained, the system can respond more timely, convenience and rapidness are achieved, the background signal value can be taken out in time, the more accurate leakage detection signal value is obtained, the influence of other gas substances in the background on the leakage detection signal is reduced, and the leakage detection accuracy is improved. And along with the increase of the detection period, the precision of the estimated derivative is also increased, so that the precision of the leak detector during leak detection is improved.

The invention discloses an end-to-end speech separation algorithm based on a speech language model, and the algorithm comprises the steps: discretizing a continuous audio into a 32-order discrete codebook sequence through a residual vector quantization coder-decoder, and introducing a transcription start symbol lt through an SOT strategy; sOSgt, SOSgt; a special separator is lt; sCgt; and a termination symbol lt; eOSgt, EOSgt; splicing a multi-person voice sequence; extracting audio depth features by using a pre-trained WavLM model, and guiding an autoregression decoder to output a separated zero-order codebook sequence in combination with a cross attention mechanism; predicting a high-order codebook sequence step by step through a non-autoregression model, configuring an independent embedding layer to fuse low-order information, and introducing a task embedding mechanism to optimize modeling; based on a special separator lt; sCgt; and slicing the multi-order discrete codebook sequence, and outputting an independent voice source through an Encodec decoder. According to the method, the intelligibility of voice separation and the audio restoration quality can be effectively improved, the decoding speed is high, the subjective hearing experiment result and the downstream task performance are excellent, the scene that the number of speakers is unknown is supported, and the industrialization application prospect is wide.

The invention discloses a power transmission positive and negative sample differential fusion method and system based on foreign matter transaction, and the method comprises the steps: collecting a standard negative sample, extracting the normal features of the negative sample, and constructing a normal feature dictionary; real-time sample images are collected, sample features are extracted in a multi-level mode and classified, and positive and negative sample feature vectors containing classification labels are output; calculating the minimum feature deviation degree between each sample feature vector and the normal feature dictionary through an anomaly measurement algorithm, and generating a sample anomaly spectrum; calculating a positive sample feature dispersion through a diversity forced separation algorithm; and constructing a differential fusion loss function according to the sample anomaly spectrum and the positive sample feature dispersion, training a foreign matter transaction detection model, and outputting a foreign matter transaction detection result. According to the method, the technical problems of high false alarm rate and weak generalization ability of the model caused by extreme imbalance of positive and negative samples in foreign matter detection of the power transmission line are solved, and high recall rate and low false alarm rate are simultaneously realized under the condition of extreme imbalance data.

The application relates to the technical field of human-computer interaction, in particular to a large-screen interaction method and system based on a handheld device, which comprises the following steps: initializing the access mode and permission configuration of a master control device, a vice control device and a large screen; adopting a double-anchor positioning recognition algorithm to distinguish and recognize gesture instructions sent to the large screen by multiple users at the same time, calculating the visual space weighted score between each gesture instruction and a candidate user, combining a trajectory integrity performance energy function to filter discontinuous and noise segments, and dynamically and credibly determining the trajectory attribution of the gesture instructions of the multiple users; and using a layer stripping gesture conflict separation algorithm to perform multi-track reconstruction and instruction source stripping determination on the visually overlapped gesture trajectories generated by the same field operation of the multiple users by the large-screen interaction control system. The application solves the problems of order confusion, gesture instruction attribution disorder and gesture trajectory overlap in the human-computer interaction of the large screen by multiple users based on the layer stripping gesture conflict separation algorithm.

The invention discloses a method for detecting measurement and control signals of an unmanned aerial vehicle in a complex network environment. The method comprises the following steps: step 1, multi-dimensional joint perception and signal representation; the method has the beneficial effects that a dual enhancement mechanism of separation and reinforcement learning dynamic anti-interference guided by physical characteristics is introduced, and normal form transformation from passive filtering to intelligent countermeasure is carried out, so that inherent robustness is improved through physical characteristic constraint on a kernel (separation algorithm); active and strategic anti-interference is realized on the periphery through reinforcement learning, and a deep defense system is formed; meanwhile, a mixed framework with unsupervised clustering pre-screening and physical feature guiding ICA as a main mode and light weight fine tuning as an auxiliary mode is adopted, fusion innovation from data driving to physical knowledge guiding and light weight intelligence is achieved, the core separation process depends on physical attributes and unsupervised learning of signals, and dependence on a large amount of labeled data is greatly reduced.