33results about How to "Ability to learn independently" patented technology

The invention discloses an intelligent vein authentication method and system with self-learning ability. In the method, the full convolutional neural network deep learning model is combined with repeated, regular global training and irregular local training. Training, before the full convolutional neural network deep learning model goes online, use the accumulated samples to conduct a global training, and save the optimized feature extraction, feature recognition and feature classification parameters; after the model goes online, use the newly registered samples to The model conducts irregular local training to complete the optimization of feature classification parameters; under certain conditions, the newly accumulated samples are used to conduct regular global training on the model, and the feature extraction, feature recognition and feature classification parameters are optimized again, ensuring that This model becomes an intelligent authentication system with self-learning ability and enhances the accuracy of vein authentication at the same time.

The invention discloses a smart home system based on deep learning. Including video recording subsystem, information processing subsystem, intelligent control subsystem, behavioral data analysis subsystem, early warning subsystem. The video recording subsystem uses high-definition cameras and machine learning to realize 24-hour uninterrupted recording of the real-time situation in the user's house; the information processing subsystem performs image data processing and feature value extraction on the recorded video data to facilitate further detection of abnormal behavior data; The convolutional neural network (CNN) algorithm detects abnormal behavior data, divides abnormal behaviors into abnormal levels, and adopts different early warning measures according to the preset abnormal threshold and abnormal level, which greatly guarantees the safety of users; the system also has behavior data The analysis function uses big data analysis technology, artificial intelligence, machine learning and other technologies to analyze the user's daily behavior data, so that the system can serve different users more humanely.

The invention relates to a prediction method of torsional vibration of a rolling mill. The prediction method comprises building a torsional vibration experiment platform and finding out an optimal prediction model of the torsional vibration of the rolling mill through real-time adoption of optimal kernel functions and parameters. The torsional vibration experiment platform is mainly formed by an alternating current frequency conversion motor, a gear reducer, a working roller which is arranged at the load end, a torsional vibration sensor, an elastic connecting shaft, a speed measuring sensor and an upper computer; the speed measuring sensor is used for measuring the speed which is transmitted to the elastic connecting shaft after the alternating current frequency conversion motor passes through the gear reducer, wherein the speed is used for representing the rolling speed of the rolling mill; the torsional vibration sensor is used for measuring the size of torque in the elastic connecting shaft; detected signals are transmitted by the two sensors to the upper computer through a serial port which is formed in the upper computer. According to the prediction method of the torsional vibration of the rolling mill, the torsional vibration prediction process is simple, the prediction accuracy is high, the objective evaluation standard of the torsional vibration is increased, the steps are clear, and the time and the labor are saved.

The invention discloses a method for classifying iris image quality based on BP network and wavelet transform, including constructing a BP neural network model for iris image quality classification; selecting several iris image samples, performing wavelet transform on them, and extracting wavelet transform coefficients The eigenvalues ​​of the extracted wavelet transform coefficients are input to the BP neural network model for training; the trained BP neural network model is used to classify the iris image quality; the wavelet coefficient eigenvalues ​​extracted from the iris image samples are input to the trained BP Neural network model; if the output result of the BP neural network model is the same as the expected value of the available iris, it is judged that the current input iris quality is available; otherwise, if the output result is the same as the expected value of the unavailable iris, it is judged that the current input iris quality is not available.