[solved] Record Audio and Response time from voicekey input

Comments

• sebastiaan

  December 2014 edited 6:59AM

  Hi,

  The soundrecorder plug-ins only save the sound files, and don't provide any way to analyze them online. To do so, you would need to write a script, such as the one from the discussion you're referring to. (Although the initialization problem with pyaudio can be annoying.)

  However, I would say it's much better to save the sound files to disk, and to determine vocal RTs afterwards during the analysis. This way you can take your time to find the optimal parameters for detecting vocal responses, because it's very difficult to do this online. Or is it important to have the vocal RTs during the experiment, for example to provide feedback?

  Cheers,
  Sebastiaan

  
  
  Check out SigmundAI.eu for our OpenSesame AI assistant!
  
• bilancino

  December 2014 edited December 2014

  Thanks for your reply.
  I'm trying to "convert" some experiments from DMDX.
  DMDX record the Vocal Rt and also the audio and then, after the experiment, you can adjust the rt according to the audio file.

  In this experiment I don't need the RT during the execution.
  So I can infer the vocal RT after my experiment is finished.

  I have two question:

  1. If I record the audio, which software I should use to extract the rt from audio file? (Audacity)
  2. I record one audio file for every loop cycle (stimuli). I have the audio file, but how can I know when (in time) my stimuli appears to measure the real response time?
     In other words, how can I start recording at the same time in which the stimulus appears?

  This is my OS structure:

  1. Experiment
  2. Loop (Setting variable)
  3. Sequence
  4. Sketchpad (Fixpoint)
  5. StartRecording
  6. Sketchpad (Stimuli)
  7. StopRecording
  8. Logger

  Thanks,
  Paolo.
• sebastiaan

  December 2014 edited 6:59AM

  Hi Paolo,

  If I record the audio, which software I should use to extract the rt from audio file? (Audacity)

  There is no standard software or procedure to do this. Most people use some custom algorithm, combined with visual inspection of the audio file. You could take a look at this paper by @lvanderlinden. It's mostly about EMG onset detection, but the problem and technique is similar for voice onset detection (the algorithm is linked to in the paper):

  • http://journal.frontiersin.org/Journal/10.3389/fpsyg.2014.01213/full

  (DMDX probably makes vocal RT detection look really easy, but, if so, that's because it uses a very crude algorithm, such as a simple loudness threshold.)

  I record one audio file for every loop cycle (stimuli). I have the audio file, but how can I know when (in time) my stimuli appears to measure the real response time? In other words, how can I start recording at the same time in which the stimulus appears?

  Recording will start where you insert the sound_start_recording plug-in. In your case, you inserted it right before the sketchpad with your stimulus, so the recording will start at stimulus presentation.

  A word of caution though: The timing of sound playback and recording can be very unreliable. This depends on many factors, mostly the specifics of your system. If you have the equipment to do so, I would test the timing beforehand.

  Cheers!
  Sebastiaan

  
  
  Check out SigmundAI.eu for our OpenSesame AI assistant!
  
• bilancino

  December 2014 edited 6:59AM

  Thanks for your advice.
  Now I have more clear some of the problem and how to solve it.
  Best Regards,
  Paolo