Approach

GSR processing in [1]

1. Propocession

• Deal with contact problems (-1 value)
• local (mechanical) disturbances : must be filtered out, because it can be mistaken as genuine peak
• apply median filter to window of step size 100
• aggregation
• discretize data with SAX (levels of Stress: 1-relexed, 5-maximum arousal) (local relative measure of arousal) (help the change detector to be more accurate)

2. Change detection :

• CUMSUM

own approach

• lowpass or median filter

• highpass to fit baseline

• or: convolve with very big filter, substract from original signal

• maybe calculate a difference function

• calculate mean magnitude of this function

ECG processing

1. remove unusable data by error value (0,5)
2. calculate beats
3. fill gaps by size
4. data from too big gaps
5. calculate HR with gaussian convolution
6. calculate HRV by diff(HR)