The data files are saved in the data folder within the Tap-It Documents.
For each user ID the following file is created:
(userID)_play_order.txt - A text file containing the actual play order. This is useful when randomizing is enabled.
For each task (audio file played) the following files are created:
(userID)_(trackNo).txt - A text file containing the timestamp data.The timestamps are in seconds.
(userID)_(trackNo).wav - A wave file of the tap recording. This file is a direct recording from the microphone of the device.
These files can be downloaded from the device using iTunes by dragging it onto your desktop or by using the "Save to..." button in the lower right corner of the main iTunes window.
Tap-It creates two types of data files, a .txt file and a .wav file. The text file contains the timestamps created by tapping the screen, while the wave file contains the microphone recording. There are pros and cons to each file. The text file is easy to read into any software or script, and it is unlikely to be a false input. However, due to the limitations of the device, there is a time latency of approximately 15ms to 20ms from a tap on the screen to the actual timestamp. (See Performance for details on the tap-to-timestamp latency.)
To compensate for the latency, Tap-It also saves the microphone recording. The recording is strictly synchronized to the playback audio, thus the taps will be exact. But, due to the fact that this is a direct microphone input, the signal is prone to ambient noise, and the microphone sensitivity varies from device to device. Silent touches may not be easy to pick up through the built-in microphones. To extract the tap onsets from the recorded audio, you will need to do further processing. A simple example code for this can be found below.
NOTE: The data files maybe overwritten if a task with the same user ID is performed. It is recommended to copy the data folder to your computer after finishing an experiment.
Matlab examples
The following example code can be used to load the data into Matlab for processing.
Functions
tapread.m - reads the timestamp data from the .txt file
findTapPeaks.m - a rudimentary algorithm with bpm assumption for finding peaks in the .wav file.
findMatchingPeaks.m - a not-so-general function for matching the detected peaks with the tap timestamps.
Examples
compareStamps.m - runnable example file for comparing the timestamps with detected peaks and plotting the errors. This was used to create the plots in the Performance section.
Performance
The tap-to-timestamp latency is approximately 15ms to 20ms with a standard deviation of approximately 5ms to 8ms, depending on the load and/or state of the device.
An iPhone 3GS and an iPod Touch 4G were used for testing. Differences between devices were negligible.
The latency was measured by comparing the peaks in the microphone recording (.wav) and the timestamps (.txt) of the same task.
Below are sample plots of the tap-to-timestamp errors. As can be seen in the histogram of errors, the distribution of the errors are complex and not necessarily Gaussian.
The recorded data overlayed with the detected peaks and the time stamps.
Zoomed-in stem plot comparing the detected peaks vs. the time stamps.
TapIt
