3 results about "Matrix decomposition algorithms" patented technology

This invention provides a method, medium, and system for acoustic early warning of thermal runaway in lithium-ion batteries based on feature reconstruction, belonging to the field of lithium-ion battery technology. The invention constructs a positive sample set by collecting safety valve opening sounds through multi-condition thermal runaway experiments, and expands the samples using data augmentation. Multi-resolution Mel spectra are extracted from the audio signals. These Mel spectra are then input into a complex-domain phase-aware separation model for joint estimation of complex-domain amplitude masking and phase residuals. Physical prior corrections are applied to the reconstruction results using a sound source separation algorithm based on wave equation time-frequency inverse scattering and a low-rank sparse time-frequency matrix decomposition algorithm based on random matrix theory. The three corrected signals are weighted and fused to obtain a corrected Mel spectra, which are finally input into a temporal convolutional network to classify and identify the safety valve opening sounds and output a thermal runaway early warning signal. This invention solves the technical problem of insufficient accuracy in thermal runaway early warning caused by acoustic feature reconstruction distortion in complex noise environments.

This invention discloses an objective fatigue state recognition method based on EEG classifier-assisted annotation, belonging to the field of human-computer interaction and biosignal processing technology. The method includes: simultaneously acquiring EEG, EMG, and eye-tracking video signals from the subject; weighting the EEG channels and extracting frequency domain features using a non-smooth, non-negative matrix factorization algorithm to construct a MobileNetV2 EEG classifier; extracting eye-tracking features using an improved Unet algorithm incorporating an attention mechanism and elliptic fitting error loss; automatically annotating the eye-tracking data using the EEG classifier and setting a threshold for fuzzy sample exclusion; and fusing the outputs of the EEG classifier and the eye-tracking sub-classifier at the decision level using D-S evidence theory to construct a comprehensive fatigue recognition model and form a closed-loop optimization paradigm. This invention solves the bottleneck problem of high-quality annotation of cross-modal fatigue data, significantly improving the objectivity and accuracy of fatigue recognition, and can be widely applied in safety-critical fields such as aviation, rail transportation, and long-distance driving.

A service soldier training and certification system includes a processing module that performs the following steps: collecting and preprocessing multimodal data to obtain skill increment feature components; establishing a digital twin roulette model, adjusting the weights of corresponding skill points in the digital twin roulette model according to the skill increment feature components and / or external feedback components, and obtaining a skill point weight matrix; obtaining a standardized course feature matrix, calculating the cosine similarity between the skill point weight matrix and the standardized course feature matrix to obtain a course content matching score; obtaining a service soldier-course cross-feature matrix and processing it based on a matrix factorization algorithm to obtain a course prediction score; linearly weighted summing to obtain the final course score, sorting the courses in the database according to the final course score, generating a course recommendation list for each registered service soldier, and pushing it out. This application can provide service soldiers with dynamic, personalized, and quantifiable training paths.