SR box problems with EyeLink

Comments

• eduard

  October 2021

  Hi Sine,

  These two problems might be related. Eyetracker and srbox should not interfere, obviously. It might be worthwhile to dig down the stack trace of sr box and check what things are done upon calling srbox.start() (either with the debugger, or modifying the source code + print statements). Perhaps, you can isolate the process that causes the interference. I don't have button box myself (and neither an eyetracker), so I can't really help you here. If you like, I can try to give you some guidance in trying to find the problem.

  Eduard

  
• SINE

  October 2021

  Hi @eduard ,

  many thanks for your reply. So, I managed to solve the first problem (it was really a silly thing - to be able to access the device srbox from the inline Python script, you need to drag the srbox plugin at the beginning of the experiment, and I had forgotten to modify the Timeout from infinite to a finite number).

  About my second question, I think you are right in saying that the problem might be linked to srbox.start(). I a while loop, where I wait for the button response and, while waiting, I check the fixation. It seems that the while loop does stop at the first iteration or simply there where I get srbox.get_button_press() - it does not even print core.getTime() a few lines below, which should set the timeout for the loop. Below you can see my loop. Perhaps there is something wrong with it and I cannot not see it (please, don't mind the ugliness of the code):

  
  

  def keyboard_response(my_canvas,target_shape,target_position,h,d,r,trigger): # SR button responses
  
  
  
  
  	eyetracker.log('searchdisplay on screen')
  
  	pupil_log = []
  
  	dev_log = []
  
  	n_gaze_samples = 6
  
  	max_gaze_deviation = angle2pix(1.5,h,d,r) 
  
  	failed_fixation_vs = 0
  
  	t0 = core.getTime()
  
  	response_made = False
  
  	buttons = []
  
  	srbox.start()
  
  	while not response_made:
  
  		
  
  		button, t1 = srbox.get_button_press(allowed_buttons=[5,6], timeout=None, require_state_change=True)
  
  		srbox.stop()
  
  		
  
  		curPos = eyetracker.sample()	
  
  		x_sample = curPos[0] - hRes/2
  
  		y_sample = (curPos[1] - (r/2))*-1
  
  		dist = np.sqrt((x_sample**2) + (y_sample**2)) # compute distance from the centre where participants need to fixate
  
  		pupil_size = eyetracker.pupil_size() # if this is 0, it means that the eyes are closed - participants blink
  
  
  
  
  		# log the pupil size. This is important for filtering eyeblinks in real-time
  
  
  
  
  		pupil_log.append(pupil_size)
  
  		pupil_logged = [i for i in pupil_log if i != 0]
  
  		median_pupil_size = np.median(np.asarray(pupil_logged))
  
  		pupil_threshold = median_pupil_size * 0.8 # update threshold for blink detection
  
  
  
  
  		# log the gaze deviation
  
  
  
  
  		dev_log.append(dist)
  
  		curr_dev = np.mean(np.asarray(dev_log[-n_gaze_samples:]))
  
  
  
  
  		if all(i >= pupil_threshold for i in pupil_logged[-9:]):
  
  
  
  
  			if curr_dev > max_gaze_deviation:
  
  				#my_canvas.clear()
  
  				fixation(my_canvas,'blue')
  
  				failed_fixation_vs += 1
  
  				print(failed_fixation_vs)
  
  				my_canvas.show()
  
  				# wait until fixation is correct
  
  
  
  
  			else:
  
  				#my_canvas.clear()
  
  				fixation(my_canvas,'black')
  
  				failed_fixation_vs += 0
  
  				my_canvas.show()
  
  
  
  
  		if (button == [5]) and (target_position[0]==-ray):
  
  			rt = t1/1000 - t0
  
  			hit=1
  
  			sent_eeg_trigger(trigger)
  
  			response_made = True
  
  		elif (button == [6]) and (target_position[0]==ray):
  
  			rt = t1/1000 - t0
  
  			hit=1
  
  			sent_eeg_trigger(trigger)
  
  			response_made = True
  
  		elif core.getTime() - t0 > 2.2:
  
  			rt = np.nan
  
  			hit=0
  
  			response_made = True
  
  		else:
  
  			hit=0
  
  			rt = np.nan
  
  			sent_eeg_trigger(trigger)
  
  			response_made = True
  
  
  
          srbox.stop()
  	search_end = core.getTime()
  
          return rt, hit, search_end, failed_fixation_vs
  

  Many thanks for your help!
• SINE

  October 2021

  Ah, while copying the code I see that I wrote srbox.stop() both inside the loop and the other at the end of the function. I had actually written it only at the end of the function in the original code (and yet, it does not work)
• SINE

  October 2021

  Hi @ eduard, I realized what the problem was, and I am not sure there is an easy way out. The problem is that the function srbox.get_button_pressed() has a timeout argument. This means that it is not possible to get a timestamp value for the srbox and loop until a response is given. The function srbox.get_button_pressed() "freezes" the loop until a response is given, without allowing me to simultaneously check for the fixation.

  Is there another version of the original function which does not have a timeout defined in it?
• eduard

  October 2021

  Hi,

  Yes, exactly, that is indeed the problem. There is a solution, and it actually isn't very difficult. Basically the trick is to have a very small timeout set (a few milliseconds), so that the loop does not freeze, but keeps on running. Then, on every iteration the srbox checks whether a response was given, and process it whenever it happens. Would that work? (I use this procedure a lot with regular keyboard presses, so I am no 100% sure it works as well for the srbox)

  
• SINE

  October 2021

  Hi @eduard,

  thanks! actually I had already tried this, but it means that the response can be given in an extremely narrow temporal window, and so it gets missed most of the times.

  Since I need very high time precision (the effect I am looking for is of the order of 20ms), this would seriously impair my data collection :(
• eduard

  October 2021

  Ah okay, this might be an issue. WIth keyboard responses, keypresses are monitored and detected all the times and saved in a buffer that Opensesame can check periodically to find a response. That way, responses usually dont get lost, even if the resp window is only 2 ms or so. However, with the srbox, it seems like there is no such buffer, or the buffer is emptied every time a response is polled. I shall try to find that clearing (in case it happens) in the code tomorrow. If there simply is no buffer for keyboard presses, an alternative would be to monkey patch the sr polling function and add eyetracking functionality to it. I don't think this should be terribly complicated.

  I'll try to have a look into it tomorrow. (but of course feel free to explore by yourself)

  Eduard

  
• eduard

  October 2021

  Hi Sine,

  Can you try to call the get_button_press function with the keyword require_state_change set to False ?

  

  That might work (if it doesn't break other things)