ingeniousfrog/deepseg

A deep learning-based motion segmentation framework for activity recognition using WiFi Channel State Information (CSI).

DeepSeg

DeepSeg aims at segmenting activities for WiFi Channel State Information (CSI)-based activity recognition.

Because fluctuation ranges of CSI amplitudes when activities occur are much larger than that when no activity presents, most existing works focus on designing threshold-based segmentation methods, which attempt to seek an optimal threshold to detect the start and end of an activity. If the fluctuation degree of CSI waveforms exceeds this threshold, an activity is considered to happen.

However, there exist some weaknesses for these threshold-based segmentation methods. First, policies of noise removal and threshold calculation are usually determined based on subjective observations and experience, and some recommended policies might even be conflicted. Second, threshold-based segmentation methods may suffer from significant performance degradation when applying to the scenario including both fine-grained and coarse-grained activities. Third, motion segmentation and activity classification, which are closely interrelated, are usually treated as two separate states.

DeepSeg tries to adopt deep learning techniques to address these problems. DeepSeg is composed of the motion segmentation algorithm and the activity classification model. The descriptions about the codes are shown as follows:

02ExtractCsiAmplitude

This is used to extract amplitudes from raw CSI *.dat files, and save as *.mat files

03DataCsiAmplitudeCut

This is used to cut the rows of data

04ManulSegmentActivity

This is used to manually mark start and end points of activities.

05ExtractSegmentTrainData

This is used to extract training data for the motion segmentation algorithm.

06DiscretizeCsiForSegment

This is used to discreize continuous CSI data into bins for segmentation

07ExtractActivitySample

This is used to extract training data for the activity classification model.

11CnnClassifyActivity

This is for training the activity classification model.

12CnnSegmentCode

This is for training the state inference model.

32FeedBackPython

This is for the joint training of the motion segmentation algorithm and the activity classification model.