Independent component analysis based glancing signal sample optimization method

Abstract

The invention discloses an independent component analysis based glancing signal sample optimization method. The independent component analysis based glancing signal sample optimization method comprises the steps of obtaining eye electric signals when a subject glances upwards, downwards, leftwards and rightwards by using eight biological electrodes; performing pre-treatment for original multi-lead glancing eye electric signals through a band-pass filter so as to remove noise interference; obtaining a spatial filter group of single glancing data in the presence of different task backgrounds by using an ICA method; performing spatial filtration for all the glancing data by using an ICA spatial filter group and performing cross test for the filtered data by using a support vector machine; and realizing sample optimization for eye movement signals by regarding an average recognition accuracy. Compared with the existing prior art, the independent component analysis based glancing signal sample optimization method has the advantages of higher recognition accuracy, stronger robustness, stronger expansibility, good application foreground and the like.

Description

technical field [0001] The invention relates to a method for optimizing samples of glance signals based on independent component analysis. Background technique [0002] Human Activity Recognition (HAR) refers to the comprehensive analysis and recognition of the observed individual's action types, behavior patterns and other information, and the recognition results are described in natural language and other ways. Since the HAR system can actively perceive user intentions, it has broad application prospects in the fields of intelligent video surveillance, medical diagnosis, motion analysis, and human-computer interaction, and has become an emerging research hotspot in the field of artificial intelligence and pattern recognition. At this stage, the acquisition of human behavior information mainly adopts two methods: non-contact environmental sensors and wearable human body information sensors. Among them, wearable bioelectric sensors can effectively make up for the shortcomin...

AI Technical Summary

Problems solved by technology

However, during the design process of the EOG-HAR system, it was found that when the acquisition process is subject to some noise interference (such as: electrode looseness, channel transient distortion, sudden interference of acquisition equipment, etc.), although the duration of this interference is short, the The signal amplitude is large, the waveform is irregular and unavoidable, so it is difficult to guarantee the recognition accuracy of the above algorithm in the actual use process, which will seriously affect the performance of the EOG-HAR system, and even the system cannot be recognized

To solve this problem, noise suppression is a relatively common solution. However, due to the strong randomness of noise types and timing, and different denoising algorithms are limited by their own performance, the effect of noise suppression in practical applications not ideal

Images