Time Functions: Settings and Flags
o Filter Method: The user can specify to filter trial data using an FFT low-pass filter or a 4th-order zero-lag Butterworth low-pass filter.
o Filter Frequency: The low-pass filter is a frequency domain (using Fast Fourier Transform), low-pass filter with bandpass equal to 1, stop band equal to 0, and overall 0-phase. In principle, it can have any filter characteristic (Teulings & Maarse, 1983), however, a symmetrical, sinusoidal attenuation band is set between Filter Frequency * (1 +- 4 / 7) while only the Filter Frequency can be selected. Refer to the section on filtering for more description.
o Sharpness factor: Can also be set to the required level in addition to the filter cut-off frequency. The implication would be that half filter width = filter frequency/sharpness.
o Read only every specified decimate-th sample: This option reads every Nth sample (where N is specified in the box) instead of reading every sample.
~This option can be used to avoid truncation of the recordings when they consist of several thousands of samples.
~If you specify decimation value of 0, MovAlyzeR will automatically determine the optimal decimation to analyze the entire recording independent of its size, by first reducing the number of samples and then by averaging into groups of 2, 3, ... samples.
~If you as the user want to determine the decimation to have a fixed value, set it to a positive value of 1 (default = no decimation), 2, 3, ..... A good start value is 1, unless truncation takes place in which case you may choose 0.
o Supress Calculation of Jerk: This option is switched ON by default, because a movement might not consist of ONLY straight strokes (e.g., it could have submovements or intentionally curved strokes).
Normalized jerk (smoothness) is only applicable in straight strokes and hence this option can be switched OFF in experiments with ONLY straight strokes.
~ Normalized jerk is defined as square-root( 1/2 * Integral (dt j2(t) * DeltaT5 / DeltaY2).
EXAMPLE: In maximally smooth movements, normalized jerk = pi 3/22 = 7.7. Smaller values are unlikely. In constant-velocity circular strokes, normalized jerk = pi 3/21.5 = 11. The increased value reflects the additional accelerations necessary to produce a curved trajectory.
o Spectrum of Raw signal instead of velocity: The spectrum can be based on the position signal instead of the velocity signal which reduces the visibility of the higher frequencies.
o Spectrum of acceleration (A_Spectrum) instead of the position or velocity spectrum:The spectrum can also be based on the acceleration, which further enhances the visibility of the higher frequencies. The acceleration spectrum is the velocity spectrum multiplied by 2*pi*frequency. Therefore the higher frequencies are becoming more visible. The unit is cm/s**2. The amplitude is the average absolute acceleration of the movement pattern.
NOTE: The above two options are mutually exclusive.
o Unrotate Automatically: Various kinds of movement patterns can be processed by the program, e.g., One or more strokes with a particular orientation <or> Drawings containing all kinds of strokes <or> Writing patterns with a particular baseline.
To properly analyze these patterns you can segment (See also segmentation) into strokes and submovements and calculate features based on the:
~ Vertical velocity, which requires that the crucial movement component is rotated to vertical (for example, longitudinal velocity in a straight line, letter height in a line of handwriting).
~ Absolute velocity, which does not require rotation.
To "unrotate" the entire recorded movement pattern to a standard direction, 2 parameters were implemented:
~ Estimate the orientation and rotate automatically to 0 degrees. The orientation is estimated by a least squares line across all samples.
~ Rotate a constant angle.
EXAMPLE 1: A long sentence will be rotated to horizontal baseline (0 degrees). Vertical velocity can then be used to segment into successive up and down strokes. The letter height is measured in the vertical direction. Vertical velocity, acceleration, and jerk represent the longitudinal component. Horizontal velocity, acceleration, and jerk represent the orthogonal component, i.e., perpendicular to the movement direction.
EXAMPLE 2: A single stroke movement. To allow segmentation using the vertical velocity, you can rotate +1.5708 radians.
EXAMPLE 3: A multi-stroke drawing with various directions may not have one preferential directory. Absolute velocity should be used to segment. Measures should be used based on absolute size, absolute velocity, absolute acceleration, and absolute jerk.
Advanced Settings
o Discontinuity Correction
When the pen is not near the tablet but lifted too high to be able to detect its movements, recording will pause until the pen is again within proximity of the tablet. This will cause a sudden jump in the data which would cause unreliable estimates of the spectrum and the time derivatives.
NOTE: Another way that the end of a recording are discarded occurs when the recording is very long while Read only every specified decimate-th sample = 1.
To salvage the trial, the user may decide to replace the discontinuity in one of several ways:
~ Discard all samples after discontinuity: By default, all data from the beginning of the discontinuity until the end of the recording will be neglected.
~ Insert constant velocity line: A constant-velocity movement is connecting the begin and end position of the discontinuity. The velocity is the average velocity. This is the second best option.
NOTE: The number of samples in the processed trial will be larger than the number of samples in the raw data.
~ Move towards start of discontinuity: The end position of the discontinuity and all following samples are moved to the beginning point of the discontinuity.
~ Accept discontinuity: Ignore discontinuities.
Several parameters are used to detect discontinuities.
o Integration constant: The intersample x and y coordinate differences are smoothed using a leaky integrator.
o Pressure of inserted or first moved sample(s): The pen pressure of the inserted part is most likely 0. Yet, the useris able to choose another value.
o Relative error: Intersample distance compared the smoothed intersample distance to detect a discontinuity.
o Absolute error: Absolute intersample distance to detect a discontinuity.