R-SMM

Codes for AAAI 2018 paper "Sensor-based Activity Recognition via Learning from Distributions".

@inproceedings{DBLP:conf/aaai/QianPM18,
 author    = {Hangwei Qian and
              Sinno Jialin Pan and
              Chunyan Miao},
 title     = {Sensor-Based Activity Recognition via Learning From Distributions},
 booktitle = {{AAAI}},
 pages     = {6262--6269},
 publisher = {{AAAI} Press},
 year      = {2018}
}

The codes are tested in Ubuntu 14.04, with Matlab R2017b.

To run the code, please follow the steps:

1. run preprocess.m to generate the corresponding RFF from activity data. The sample data is from Skoda dataset. Notice the rsmmWrite function is generated from folder .\rsmmWrite. If you run the algorithm in a different OS, then you will need to run make.m first.
2. run rsmm.sh in terminal to train and test the proposed R-SMM algorithm. It will take around 5 minutes. If you apply the proposed method to other applications, please tune the parameters. The prediction results of each activity will be stored in test.txt.