The project is based on ongoing research project MT-Diet at iMPACT Lab arizona state university. This project was done as part of CSE 572 Data Mining course under Prof. Ayan Banerjee.
This project is carried out in 5 phases as explained below.
For creating dataset we have recorded video of eating person with wristband sensors. We have used Myo armband, wearable gesture control and motion control device for capturing motion activity.
The food is equally divided into four sections of a plate. The unit of eating actions is considered as one bite. The dataset was created consisting of 40 bites each with fork and spoon.
While start eating eating we did a unique gesture that can be easily identified in the accelerometer sensor so that later we can synchronize the time stamps of the video and the accelerometer data from the wristband.
In this phase we have annotated the collected raw data as eating or non-eating with the help of video recording. We have annoted frame numbers as starting of eating activity and ending of eating activity. We captured all these frames numbers in Annotation.txt
To extract feature from raw sensor data, we have applied certain transformation on raw data & plotted graph of eating and non-eating actions. The transformations which give clear distinction between eating and non-eating actions are selected as features. We have applied following methodologies to extract the features
- Root Mean Square (R.M.S)
- Fourier transform
- Energy/Power of Signal
- Statistical Features like mean,std,max,min
TODO
TODO
- Prashant Gonarkar (pgonarka@asu.edu)
- Tanmay Manolkar (tmanolka@asu.edu)
- Pooja Koasla (pkosala@asu.edu)
- Shubham Pathak (spathak6@asu.edu)
This project is licensed under the MIT License - see the LICENSE.md file for details