Mobile terminal multi-sensor based behavior recognition model training method and device

Abstract

The invention provides a mobile terminal multi-sensor based behavior recognition model training method and device. The method includes: step 1, collecting raw data of multiple sensors on the mobile terminal according to a preset sampling frequency, and classifying all the raw data according to the behavior patterns of the collection objects to obtain sampling data sets of different behavior patterns; step 2, for Sampling data sets for each behavior pattern, compare the original data of each sensor at adjacent sampling moments, and determine multiple sets of feature vectors for each behavior pattern; Step 3, using the improved Markov chain assumption or naive Bayesian classifier Probabilistic statistics are performed on multiple sets of feature vectors of each behavior pattern, and the feature vector with the highest probability in each behavior pattern is used as the behavior recognition vector of the behavior pattern. The invention improves the utilization rate of data, reduces the complexity of the identification process, and the identified behavior content is more specific and closer to reality, has higher occurrence frequency and stronger application significance.

Description

technical field [0001] The invention relates to the technical field of data analysis, in particular to a method and device for training a behavior recognition model based on multiple sensors of a mobile terminal. Background technique [0002] In earlier studies, researchers placed special motion sensors on different body parts of the participants, such as waists and wrists, to store behavioral information data and convert them for analysis and identification. Great impact, unable to carry out normal work and life. Therefore, the above research methods cannot provide a long-term effective solution for activity monitoring or prediction. Subsequently, the continuous development of cameras has made video recording an important means in this field. Researchers use cameras to record the behavior of the observed person, and can record detailed parts of the behavior data according to the needs. This method is similar to obtaining data from sensors attached to the body. Compared wi...

AI Technical Summary

Problems solved by technology

[0004] However, in the current research on behavior recognition using mobile phone sensor data, most of the research uses a small number of sensors, a single data type, and the identifiable behavior content is simple and the algorithm is difficult to implement. The feasibility needs to be improved.

Images