Dymnz

- [x] 4 page - [x] Zip and upload to Google drive - [x] CH/EN thesis (word & pdf) - [ ] Meeting prez. (pdf/ppt) - [x] Code * Including...

Thesis Revision

1

- [x] 圖片標題改為冒號 - [x] Citation Title Case - [ ] Multiple in text citation format not changed * **Change after all else is finished to not interfere w/ zotero**...

Additional Experiment

2

* 4 more subject * 4-ch x4, 6-ch x3. 7 sessions * 30mins per session, 3.5hrs * Electrode * 4x4x2 + 6x3x2 + 7 = 82 electrode * 82 x...

Wrist Angle

7

[Reports](https://github.com/dymnz/sEMG/tree/master/Reports/wang/semg_angle_est)

2 separate NN decoding FLX/EXT or PRO/SUP. FLX/EXT-nn should have zero output when PRO/SUP-nn, vice versa. - [x] Check IMU angle, how Roll/Pitch reacts to FLX/EXT, PRO/SUP and FLX/EXT+PRO/SUP *...

* SPS: Downsample ratio should be a integer, as the decimation doesn't require upsample * w/ 2660Hz sampling rate, using * 95 SPS for high * 35 SPS for low...

Quality of nICA affects the performance of LSTM

1

* Check 1S2WA_23_CV` * W/ nICA_2 as demixing baseline, the accuracy isn't better than RMSDown * Noted in `Meeting 2018/08/30 - S2WA_23`

In S2WA_23, PRO/SUP error is higher than S2WA_22 (20/16 v.s. 8/7). FLX/EXT error remains the same. This may be caused by the way median value for 4 gesture is calculated...

Roadmap

1

- [x] RNN test * [Note](https://github.com/dymnz/sEMG/blob/master/RNN/RNN_sim.md) - [x] Vanilla - [x] Convolution test - [x] LSTM - [x] Sensor fusion test - [x] Cross-validation w/ patience - [ ] 4-channel...