Parkinson's Measurement     Home         PDMeasure         Postings         Data         Studies         Tools         Epidemiology         Philosophy         Contact        
Research by PwP for PwP   NEW... PDMeasure: integrated survey and test environments.        See.         NEW... Levodopa equivalent pharmacokinetics minute by minute.        See.        

Tremor Test


This test aims to estimate the size and frequency of a tremor.

It requires a computer with a touchpad, touch screen or mouse which is sensitive enough to detect your tremor. To see if this is the case, try this now: put your finger lightly on the touchpad, or your hand on the mouse, letting it move with the tremor. Does the cursor move? If not, the results will be inaccurate (but read on: we describe alternatives in the side panel.)




Settings To run the test, press the blue button below.
Keep a finger on the touchpad. After the set time
the program will show the results.


test duration (sec)
x calibration factor
y calibration factor
No response timeout (ms)



As used here, x refers to horizontal and y to vertical movements of the pointer on the screen, and to the corresponding left/right (x) and forward/backward (y) movements of the finger on the touchpad.


Do I need to calibrate? Only if you want to know the distance moved in millimeters, as opposed to pixels. If you are only interested in the frequency of the tremor, there is no need to calibrate.

How do I calibrate? The calibration factor will vary from machine to machine. It should be set to the number of screen pixels that the pointer moves following a 1mm movement on the touchpad. (Decimal values are allowed.) One way to do this is, with the calibration factors set to 1, run the test and just move the finger once across the touchpad then wait for the test to finish and note the x movement (call this X), then measure in millimeters the width of the pad (call this W). The x factor is then X/W. A similar thing can be done for y.

What is the "No response timeout" setting used for?

It is used as a filter in the frequency analysis. During the test there may be periods of non-tremor in which the finger moves intermittently causing rogue low frequency cycles to be detected. These, if left uncorrected, will distort the frequency values. The no response timeout value is used to void a potential cycle if the time between movement events is longer than this.

How to get the best out of the program using a touch pad

The finger under test, should rest lightly on the touchpad, facing in the y direction at about 45 degrees to the vertical. The hand and the arm should not be in contact with either the machine or the desk. If the pointer movements on the screen do not reflect the finger movements, here are a few things to try:
- try to "float" the finger; avoid it pressing down heavily;
- clean the pad to decrease friction;
- use a small polished washer between the finger and the pad;
- turn off any pointer acceleration.

How to get the best out of the program using a mouse

Everything else being equal, the lighter the mouse is the better.

Understanding the output

The average movements are calculated by adding together the absolute movements of each step and dividing by the elapsed time. It is not the same as the net distance moved per second.

A pointer movement may involve both an horizontal and a vertical pixel change. The "diagonal" measure is calculated from the sum of the straight line distances of the individual steps.

The average frequency is calculated in two ways:
- as the reciprocal of the mean period;
- as the reciprocal of the median period.
If these diverge to an extent which concerns you, the test should be repeated.

After a run, if you want more details, press the Trace button to see the raw data that has been collected and the on-going analysis done on it.


We measure movement in two dimensions even though it can have components in three dimensions.

We measure the tremor over a brief period of time. While for many people their tremor varies over time. In which case, the snap shot we make does not tell the whole story.

The treatment of frequency is heuristic: it is assumed that there is one major frequency, with the differing cycle lengths found while analysing the trace being caused by sampling error. In fact, there may be a number of different, superimposed frequencies. In which case a spectral approach would be better.

For the test to be accurate, the movements induced by the tremor must be mirrored linearly by the movement of the pointer on the screen. Put simply, if the finger moves, the pointer should move in the same direction; and if the finger moves twice as far the cursor should move twice as far. Care should be taken to avoid the pointer going over parts of the screen that change the arrow pointer symbol. All this must be done without affecting the behaviour of the tremor. How to achieve this will vary from machine to machine.

If the touchpad is not sensitive enough to detect the tremor or, conversely, the tremor is so large that the cursor is continually being taken off the screen, the results will be inaccurate.

The part of the body with the tremor may be impossible to place on the touchpad. For instance, it is difficult measuring a foot tremor in this way.

An alternative approach using Audacity

A program called "Audacity" can be downloaded for free from
Amongst other things, it allows you to make sound tracks and visualize them with precise timing.

To find the frequency of a tremor all you need to do is to let the tremor repeatedly tap a noisy surface: a table, for instance. The noise this generates is picked up by the computer's microphone and "drawn" on the screen. The taps are usually clearly visible over the background noise. All that needs doing then is to count the taps over a short period. Normalizing this to one second will give the frequency in hertz.

Audacity is also able to run a spectral analysis which is useful if there are superimposed tremors.

Unfortunately, it is difficult to see how Audacity can be used to measure the magnitude of tremors.