- Junhyun Kim @junbread
- Hosung Ryu @chrominium97
- Detect human sleeping habits using WiFi signals
- none: 사람 없음
- still: 정자세로 가만히 누워 있음
- crouch: 새우잠
- flipped: 엎드려 잠
- rolling: 뒤척임
<collector>
+
| Emits
v
+----------------------+
|Binary CSI data (.dat)|
+----------------------+
+
| For each tag,
<preprocessor> 1) Merge .dat with same tag
2) Export them into single .npy
|
v
+---------------------------------+
|NumPy arrays (.npy) |
| |
|Amplitude and optional phase data|
| |
|Shape: (pkts x 30) without phase |
| (pkts x 60) with phase |
+---------------------------------+
+--------------------+
| v
| Merge .npy files with same prefix
| <classifier/data> and split them into training and testing dataset (.pkl)
|
| |
| v
|
| +-----------------------------------+
| |Pickle exports (.pkl) |
| | |
| |Test and training dataset |
| | |
| |Test dataset: dataset.test.pkl |
| |Training dataset: dataset.train.pkl|
| +-----------------------------------+
|
| +
| |
| Train network from given dataset
| <classifier/train> and save model
|
| |
| v
|
| +-----------------+
| |Keras model (.h5)|
| +-----------------+
|
| +
| |
|
+-------> <classifier/predict>
|
v
+-----------------------------------+
|List of predictions for each packet|
+-----------------------------------+still 4h 0400 4 * 3600 * 500 7200000
crouch 2h 0600 2 * 3600 * 500 3600000
rolling 2m 0602 2 * 60 * 500 60000
flip 30m 0632 30 * 60 * 500 900000
rolling 1m 0633 1 * 60 * 500 30000
still 1h 0733 1 * 3600 * 500 1800000
rolling 2m 0735 2 * 60 * 500 60000
still 25m 0800 25 * 60 * 500 750000