sam81 / pychoacoustics Goto Github PK

ABX process and process results text

The response box could have a panel showing the results update after each trial. The plot could be a staircase for adaptive procedures, and a barplot showing percent correct and/or d' for constant procedures.

Tests need to be written for many functions/procedures:

• paradigms
• pysdt
• sndlib

Smileys are one of the feedback options for the response light box. They're not displaying on OS X (where also application icons are not displaying).

It may be good to automate the process of adding tags (e.g. beta) to the builds

Hi,

It appears I cannot run pychoacoustics with the version if I clone from github.

Here's what I did:

1. Unzip "pychoacoustics-master.zip" as downloaded from Github.
2. In command prompt, python setup.py install.
3. In command prompt, python pychoacoustics.pyw.
4. I get an error as written below:

Traceback (most recent call last):
File "pychoacoustics.pyw", line 38, in
from pychoacoustics import qrc_resources
ImportError: cannot import name 'qrc_resources' from 'pychoacoustics' (C:\Users\20185408\surfdrive\Spyder\pychoacoustics-master\pychoacoustics_init_.py)

I tried to compile the qrc resources myself and did the following steps:

1. To compile, I use:
   pyrcc5 -o qrc_resources.py resources.qrc
2. I copied qrc_resources.py to /build/lib/pychoacoustics/
3. Rerun python setup.py install.
4. I get the same error.

How can I get this fixed?

Add support (translations, fromatting of numbers and dates) for languages other than English in the control window.

• all the functions printing numbers '{0:5.3f}'.format(myFloat) should use self.currLocale.toString()
• the sndlib and other modules should use translations too

Hi,

I can't seem to load a prm file again in Windows. Here's the steps to reproduce it.

1. Unzip "pychoacoustics-0.5.10.zip" as downloaded from the website.
2. In command prompt, python setup.py install.
3. In command prompt, python pychoacoustics.pyw.
4. Click "load prm", and select a previously saved prm file.
5. I get the error below.

Traceback (most recent call last):
File "C:\Users\20185408\surfdrive\Spyder\pychoacoustics-0.5.10_myV2\pychoacoustics\control_window.py", line 4947, in onClickLoadParametersButton
self.loadParameters(fName)
File "C:\Users\20185408\surfdrive\Spyder\pychoacoustics-0.5.10_myV2\pychoacoustics\control_window.py", line 5187, in loadParameters
tmp['b'+str(blockNumber)]['fileChooserCheckBox'].append(strToBoolean(allLines[foo['b'+str(blockNumber)]['startFileChooser']+i].split(':')[2].strip()))
File "C:\Users\20185408\surfdrive\Spyder\pychoacoustics-0.5.10_myV2\pychoacoustics\utils_general.py", line 68, in strToBoolean
return out
UnboundLocalError: local variable 'out' referenced before assignment

Here's what I have tried:

1. Uninstall python completely, restart and reinstall it. I still get the same error.
2. Clone the current version on github - that gives me another issue.
3. Run Pychoacoustics in Linux in a VM. PRM file can be loaded successfully.

Is this a Windows specific problem? How can I resolve this?

Compress files before sending by e-mail

Have a comprehensive transformed up-down procedure in which the step size can change more than once for the initial turnpoints. The procedure should also allow that the track be terminated when either a fixed number of turnpoints, or a fixed number of trials has been reached.

Add 'Delete All' button. This may be useful in some cases. But its utility must be weighed with the downside of adding another button which may make the interface less useful. An alternative may be to have a menu entry in the toolbar.

Add vocal instructions at beginning end of each block.

Add plots for all paradigms.

The logging module is already partly used to log error messages, but its use should be extended, also check errors on Windows with python GUI mode.

It would be good to have a commandline option to start pychoacoustics with all windows hidden. This would be useful for running simulations in the background. Currently the best alternative is the commandline option to start with minimized windows.

The block-summary and session-summary result files should be named with suffixes according to this categorization. The "full" files could be renames as "trial-summary" result files.

Several points:

• the name "condition" should be reserved to the experimental condition labels assigned by the user to the experimental conditions. For example result files should peraphs use "trial type" for differentiating between a "same" and a "different" trial in a same-different experiment. Likewise, in experiment with multiple constant stimuli, the "conditions" within a block should be consistently referred to as sub-conditions (this is already done for some paradigms).
• use of underscore or camel case should be consistent
• the nDifferences parameter should be renamed nSubconditions (this works for both number of adaptive tracks, and number of constant differences)

Documentation is currently incomplete. The pychoacoustics manual needs to cover all functions.

Document

• Installation (mostly done)
• Graphical user interface (mostly done)
  • Write PSI paradigm widgets
  • Review UML paradigm widgets
• Command line user interface (mostly done)
• Paradigms (a lot to do)
• Result file (partly done)
• Default experiments (partly done)
• Engine (partly done)
• Writing your own experiments (a lot to do)
• sndlib
• pysdt

Edited unsaved fields and choosers could change color (use on editing finished)

Add d' calculation widgets.

Add PySoundFile https://github.com/bastibe/PySoundFile as wav manager option.

Add a dialog to fit psychometric function from full_table files.

The full files could record the start of each sound and the time the listener responded

on pyqt5 a QMainWindow without a parent is a primary window and has an entry in the taskbar, but a
QMainWindow with a parent is a secondary window and doesn't have an entry in the task bar (see http://doc.qt.io/qt-5/application-windows.html). The response box, being a child of the control window doesn't have an entry in the task bar and cannot be shown independently of the control window, which is annoying.

they should be called something like "signal+noise" and "noise"

Write 'what's this' and tooltips for all widgets. The 'what's this' are currently written for the text fields, choosers etc. Copy them also for the text field labels, chooser labels, etc...

Add block performance feedback at the end of each block. Usually it's not a very good idea to give block-level feedback because participants should not know about different experimental conditions. However, block-level feedback be added as an option.

Block level and session level fields/choosers could have different colors or some other distinguishing mark.

Currently the UML procedure can be used only with n-Down n-Up sweetpoint selection rules. The ability to run the UML procedure with a random sweetpoint selection rule should be added.

The end of session command is not executed if results file name contains spaces.

Generate html report through intermediate rst file.

If a plain text or tabular result file can't be processed (because no procedure exists to process it) from the dialog process results, a warning (e.g. a pop-up message) should be given to the user

Could experiment on using a separate QThread for playing sounds in trials. Should make sure that this doesn't compromise the synchronization of the sounds with the lights.

Correct/Incorrect and smiley feedback options. Also, add the ability to change color of feedback lights.

The lambda estimation point is fixed at a level where performance correct is expected to be at ceiling. The choice of this value may be problematic in some situations (e.g. finding the level of a tone that works well for normal hearing and hearing impaired listeners in a signal detection task may be tricky). To address this issue there should be an option to change the lambda estimation point dynamically.

The proposed change is detailed below:

use the inverse of the chosen psychometric function to estimate the stimulus level at a high percent correct point (e.g 99%) on the psychometric function using the current parameter estimate:

xCeiling = invPsy(0.99-estLambda, estAlpha, estBeta, gamma, estLambda)

and use this stimulus value for the lambda sweetpoint when it exceeds a 'suggested' stimulus value that should work for most listeners:

stimValForLambdaSwtp = max(suggestedVal, xCeiling)

The procedure would use this value for estimating lambda, except if it falls below the estimated stimulus value at the 0.99-estLambda proportion correct. As a final touch one could add an absolute upper stimulus level 'highestStimAllowed' (dictated by equipment limitations or safety concerns) that will be used if the 0.99-estLambda value is very high:

stimValForLambdaSwtp = min(max(suggestedVal, xCeiling), highestStimAllowed)

Replace os.system with subprocess, possibly popen for sending kill signal. This has been partly done but need to check all code.