A method and system for processing outliers in electromyographic data

Abstract

The present invention relates to a method and system for processing outliers in electromyographic data, comprising: obtaining a data set of data samples, and dividing the data set into a first-type positive domain, a first-type boundary area, a second-type positive domain, and a second-type positive domain The second type of boundary area; clustering the data samples in the first type of boundary area and the second type of boundary area, and according to the clustering results, respectively dividing the first type of boundary area and the second type of boundary area into multiple groups; According to the distance from the data samples in the first type of boundary area to the first type of positive domain and the distance from the data samples in the second type of boundary area to the second type of positive domain, determine the distance between each group in the first type of boundary area and the first type of positive domain. The average distance of the domain and the average distance from each group of the second type of boundary area to the second type of positive domain, the group whose average distance is greater than the distance threshold is marked as an outlier point, and the set of outlier points is used as an outlier set; the data Outliers whose sample size is greater than the enclave threshold are marked as enclave groups, and the enclave groups in the outlier set are deleted to obtain the outlier processing results.

Description

[0001] Technical field [0002] The invention belongs to the field of machine learning, and in particular relates to a method and system for processing outliers in electromyographic data, which is applicable to the scene of using electromyographic signals to detect abnormal muscles. Background technique [0003] When using EMG signals for abnormal muscle detection, a large number of EMG samples with labels (that is, abnormal muscles and normal muscles) are often required, and an abnormal muscle detection model is trained based on this data set to achieve the purpose of identifying abnormal muscles . It is hoped that the abnormal muscle recognition model can achieve a good generalization effect on the real data set in the future, but the generalization ability of the model is closely related to the quality of the data set. Due to various external influences in the process of EMG signal acquisition, such as motion interference, electromagnetic interference, power frequency nois...

AI Technical Summary

Problems solved by technology

[0005] In order to solve the problem of the influence of outliers in the data set on the generalization ability of the model in the process of establishing the abnormal muscle recognition model, the present invention mainly solves the problem that the existing outlier point recognition method misidentifies the outliers and non-noise groups caused by insufficient samples in the data set The problem that the group is mistakenly identified as an outlier and deleted, and it also solves the problem that the existing method cannot identify all the outlier sample groups with noise patterns

Images