% Voice onset extraction
1. Smoothed amplitude (SA) is calculated by moving average of abs(signal) 
   with time window length of t_smooth (10ms : 100Hz)
2. Gamma distribution is fitted to the SA histogram 
3. Default threshold value (y0) is determined from estimated gamma distribution
4. Threshold value yh = y0 * status_level
5. Voice onsets are extracted where SA exceed the threshold

--- Condition for voice onset (vb_get_voice_onset.m)
1. Period length that SA exceed the threshold is larger than 't_period'

--- Optional parameter for voice onset                              [Default]
parm.t_period : minimum period length that SA exceeds the threshold [100 ms]
parm.status_level : threshold value control rate                    [1.0]

--- Usually following parameters need not be changed
parm.t_smooth : moving average window length                        [10 ms]
parm.p_val    : P-value corresponding to the threshold              [0.0001]
     If the signal exceeds the threshold, 
     the probability, that the signal is noise, is less than p_val

% EMG onset extraction
1. Smoothed EMG is calculated by moving average of abs(EMG) 
   with time window length of 't_smooth' (20ms)
2. Gamma distribution is fitted to the peak mode of EMG histogram 
3. Default threshold value (y0) is determined from estimated gamma distribution
4. Threshold value yh = y0 * status_level
5. EMG power is calculated by TKE operator
	1) band-pass filtering at 30-300 Hz
	2) y(t) = x(t)^2 - x(t-1)x(t+1)     (TKE operator)
	3) low-pass filtering at 50 Hz 
6. EMG onsets are extracted where EMG power exceed the threshold value

--- Condition for EMG onset (vb_get_emg_onset_time.m)
1. Period length that EMG power exceed the threshold is larger than 't_period'
2. Distance from previous onset should be larger than 't_event'

--- Optional parameter for EMG onset                         [Default]
parm.t_event  : minimum distance from previous onset event   [500 ms]
parm.t_period : minimum period that Smoothed EMG > threshold [50 ms]
parm.status_level : threshold value control rate             [1.0]

--- Suggestion  for optional parameters
* If too many EMG onsets are extracted, try to increase 't_event'
* If some of onsets are not extracted, 
     try smaller status_level and/or shorter 't_period'
* If false onsets with small peaks are extracted,
     try larger status_level and/or longer 't_period'

--- Usually following parameters need not be changed
parm.t_smooth : moving average window length                 [20 ms]
parm.p_val    : P-value corresponding to threshold value     [0.001] 
     If the signal exceeds the threshold, 
     the probability, that the signal is noise, is less than p_val

