8 results about "Blind source separation algorithm" patented technology

The application discloses a machine vision assisted automobile injection molding part size precision monitoring method and relates to the technical field of machine vision, which comprises the following steps: combining visible light and near-infrared light to irradiate an injection molding part, synchronously collecting three-view images, segmenting the injection molding part region and framing the ROI region of a key size, and fusing to generate a multi-view fusion image; identifying the smallest key feature size from the ROI region, decomposing the fusion image through multi-scale morphological iteration processing, extracting a real contour signal by using an ICA blind source separation algorithm, and generating a complete contour model through multi-view edge point fusion reconstruction; marking feature points based on a CAD standard model, positioning the feature points by using a particle swarm optimization and a Bayesian iteration algorithm, and calculating key size parameters; collecting environmental and injection molding part surface error factors, constructing a dynamic error calibration model to compensate for size deviation, and outputting the final size value after calibration, so that the accuracy and reliability of automobile injection molding part size detection are significantly improved.

This invention proposes a method for suppressing short-duration diurnal magnetic field interference based on dual-aircraft collaboration. Addressing the challenges of establishing fixed diurnal variation monitoring stations and lacking effective diurnal variation reference information in offshore dual-aircraft collaborative airborne magnetic surveys, this method utilizes airborne magnetic survey data collected during near-shore airborne magnetic surveys and synchronous observation data from diurnal variation monitoring stations. It proposes a dual-aircraft collaborative diurnal magnetic interference suppression method based on a neural network-guided blind source separation algorithm. This method first uses a deep neural network to learn and extract prior features of diurnal variation interference from the dual-aircraft observation signals. These features are then used as guiding information in the blind source separation process to separate the independent source components corresponding to the diurnal magnetic interference, ultimately achieving high-precision identification and suppression of diurnal variation interference components in the original signal. This method can efficiently separate and suppress short-duration diurnal magnetic interference, significantly improve the signal-to-noise ratio of magnetic survey data, and effectively preserve the target magnetic anomaly characteristics, providing reliable technical support for high-precision airborne magnetic surveys.

This invention discloses an in-vehicle selective noise reduction method based on sound field reconstruction, relating to the field of in-vehicle safety warning technology. The method includes: collaboratively collecting in-vehicle acoustic and vibration data through sensors; separating the acoustic and vibration data using a blind source separation algorithm to obtain a voiceprint feature tensor; inputting the voiceprint feature tensor into a convolutional neural network for voiceprint recognition, outputting a safety sound category and a safety sound hazard level; generating adaptive noise reduction parameters based on vehicle speed and safety sound category; assigning the final adaptive noise reduction parameters to seats in different regions; and issuing a physical vibration warning based on the safety sound hazard level to complete the in-vehicle selective noise reduction. This invention significantly improves the detection and recognition accuracy of critical safety sounds in road noise environments; it balances noise suppression and active safety assurance, effectively enhancing the vehicle's perception and response capabilities to emergency sound sources while improving passenger comfort.

The application relates to the technical field of brain-computer interface, and discloses a signal acquisition method for brain-computer interface control of a lower-limb exoskeleton robot. The method collects original electroencephalogram signals through a multi-channel electrode array arranged on a motor cortex of a scalp, and synchronously collects motion data of a lower-limb inertial measurement unit. In the signal processing process, differential amplification is adopted to suppress common-mode interference, adaptive narrow-band filtering is adopted to dynamically track electroencephalogram rhythms, and a blind source separation algorithm based on independent component analysis is adopted to remove motion artifacts. Finally, the processed motion intention electroencephalogram signals are output, the precision and real-time performance of intention recognition are improved, and accurate and stable control signals are provided for the lower-limb exoskeleton robot.

The application discloses an intelligent sleep staging and fatigue assessment system based on a multi-modal deep neural network, relates to the technical field of intelligent sleep, and comprises a cECG data acquisition module, a multi-modal decoupling filtering module, a cross-modal quality calibration module, a multi-modal deep neural network algorithm module and a result output layer. The cECG data acquisition module adopts an array capacitive sensing layout, is embedded into a daily sleep carrier, and continuously acquires original array cECG signals during user sleep. Through a space and mode double-constraint blind source separation algorithm, accurate decoupling of three types of mode signals, i.e., electrocardiogram, respiration and body movement, is realized, cross-modal quality calibration and dynamic fusion weight adjustment are combined, the signal quality and fusion accuracy are improved, and with the aid of a space, quality and attention three-dimensional mutual feedback calibration network and an adaptive state gate Mamba, synchronous and accurate execution of double tasks of sleep staging and fatigue assessment is realized.

The present application relates to physiological micro-vibration signal processing technical field, specifically disclose physiological micro-vibration signal multi-dimensional time sequence feature extraction and reconstruction method and system, the method includes preprocessing physiological micro-vibration original mixed signal, obtain pretreatment signal;Based on the improved adaptive variational mode decomposition algorithm and blind source separation algorithm to the pretreatment signal processing, obtain time-frequency domain feature vector;The method of combining bidirectional long short term memory network and encoder is used to the pretreatment signal deep time sequence feature extraction and dimension reduction, obtain dimension reduction feature vector;Time-frequency domain feature vector and dimension reduction feature vector are fused, effective fusion feature vector is screened based on correlation threshold value, physiological micro-vibration signal is reconstructed based on the effective fusion feature vector screened out. Through the improved VMD-BSS algorithm, the adaptive determination of the number of modes and the penalty factor is realized, the dependence on prior knowledge is reduced, and the processing speed of signal decomposition is significantly improved.

This invention belongs to the field of online detection technology for metal impurities in food, and particularly relates to an intelligent sensing-based online detection method and system for metal impurities in food. It simultaneously transmits and receives multiple discrete frequency band electromagnetic signals, generating a multi-dimensional original signal matrix by leveraging the differences in their responses to metals, food, packaging, and the environment. Based on the fusion application of an improved FastICA blind source separation algorithm and wavelet packet transform, a multi-frequency feature-blind source separation dual-layer model is constructed. The effectiveness of signal separation is verified using a multi-scenario standard signal feature library and cosine similarity matching. After feature extraction and combined dimensionality reduction processing, the output is fused through a hybrid model of traditional machine learning and lightweight deep learning. The D-S evidence theory is introduced and combined with real-time parameters from the production line to dynamically adjust the decision rules.

This application provides a distributed optical fiber monitoring method and system for tunnel support structures, belonging to the field of safety monitoring technology for carbon tunnel engineering. The method involves adapting and deploying multi-core sensing optical cables to the support structure, with the monitoring section aligned with the stress-sensitive area, and the free section fitted with a protective sleeve that is detachably connected. A frequency-agile optical comb detection sequence is transmitted through an optical frequency comb detection module to acquire full-lifecycle sensing signals. Wavelet threshold denoising and blind source separation algorithms are used to process the signals, constructing a two-dimensional feature map monitoring dataset. An improved neural network model containing an SPPF spatial pyramid pooling layer and a path aggregation network is input to achieve accurate identification and location of cracks, anchor bolt loosening, and surrounding rock loosening, as well as anomaly classification. An early warning information feedback terminal is generated, and the detection parameters and monitoring sensitivity are adaptively adjusted.