our paper[https://ieeexplore.ieee.org/document/8662651]

if the 3-axes of the human motion model are considered as the 3 channels of a RGB image, the value of the XYZ axial data can be mapped into the value of the RGB channel data in a RGB image respectively. Namely, each 3-axial data can be converted into an RGB pixel. The 400 pieces of 3-axial data cached in the sliding window can be viewed as a bitmap with size of 20 or 20 pixels.

you can use .utils.transform.data2image func to make sensor data to img

ADLs and fall data graph

sensor data to img

we use imgs to train our network

• accuracy = 0.978718

• python3
• tensorflow 1.4.0
• pandas
• numpy
• matplotlib

python ./src/cnn.py

we need two public datasets.

• 1.MobiFall＆MobiAct DataSet http://www.bmi.teicrete.gr/index.php/research/mobiact

• 2.SisFall http://sistemic.udea.edu.co/en/investigacion/proyectos/english-falls/