(a) Typical and (b) distorted eye image with a detected pupil and corneal reflection. The images are cropped to show only the eye region. 

Contexts in source publication

Context 1

... circuit diagram that presents the coupling of the capacitances in the measurement is illustrated in Figure 3.2. The capacitance is measured between the transmitter and the virtually grounded receiver. The capacitances C x and C si are changing due to the facial movements. The device ground is connected to the target via the capacitance C g that is formed when the head-mounted prototype device is worn. Changes in this capacitance contribute to offsets in the measurement output (Smith, 1996). Figure 3.2: A circuit diagram of the coupling of the measurement in the two- electrode mode. Cx is the capacitance that is formed between a transmitter (TX) and a receiver (RX) electrode. Csi are shunt capacitances between the electrodes and the target. Cp is the parasitic capacitance between the transmitter and the device ground (GND). Cg is the capacitance between the grounded target and the device ground. The virtual ground connection between the receiver and the device ground is shown as a short circuit. Vs represents the excitation voltage source of the ...

Context 2

... circuit diagram that presents the coupling of the capacitances in the measurement is illustrated in Figure 3.2. The capacitance is measured between the transmitter and the virtually grounded receiver. The capacitances C x and C si are changing due to the facial movements. The device ground is connected to the target via the capacitance C g that is formed when the head-mounted prototype device is worn. Changes in this capacitance contribute to offsets in the measurement output (Smith, 1996). Figure 3.2: A circuit diagram of the coupling of the measurement in the two- electrode mode. Cx is the capacitance that is formed between a transmitter (TX) and a receiver (RX) electrode. Csi are shunt capacitances between the electrodes and the target. Cp is the parasitic capacitance between the transmitter and the device ground (GND). Cg is the capacitance between the grounded target and the device ground. The virtual ground connection between the receiver and the device ground is shown as a short circuit. Vs represents the excitation voltage source of the ...

Context 3

... of the transmitting and receiving electrodes for the capacitance measurement was similar to that in capacitive feather-touch buttons. Electrodes were made on a printed circuit board as a planar structure in a coaxial layout with three concentric rings. Innermost and outer- most electrode rings were coupled together forming the transmitter. Middlemost electrode ring functions as the receiver. The circuit-board layout of the electrodes is shown in Fig. 3. Figure 2 shows the capacitance controller on the small circuit board placed on the frames between the eyes. The electrode circuit board was placed directly behind the controller circuit board forming a unit for capacitance measurement. The placement of the unit was decided based on preliminary measurements carried out with different electrode ...

Context 4

... example of a signal registered during an experiment with indicated and detected facial movements is shown in Figure 3 The results of the study were that (87.7 ± 17.7)% of the frowns and (97.0 ± 5.0)% of the eyebrow raises were detected. However, some of the frowns were incorrectly detected as eyebrow raises. Only (66.1 ± 41.9)% of the frowns were correctly detected whereas a correct detection of (91.7 ± 11.0)% of the eyebrow raises was achieved. The values are means and standard devi- ations taken over all the repetitions of the experimental procedure. A single participant performed the procedure multiple ...

Context 5

... test procedure consisted of measuring the capaci- tance signal while the test subjects were indicated to frown and lift eyebrows. Indications were carried out by playing audio samples to the subjects. The audio samples were the words ''frown'' and ''lift'' produced by a speech synthe- sizer. A single test procedure consisted of a total of 25 indications, each randomly chosen to be either of the two possibilities. Intervals between the indications were chosen to vary randomly between 2 and 10 s. The test procedure was repeated four times with each subject. Test subjects were given the same simple instructions as in the brief experimentation phase. They were also instructed not to rush with their movements, but to calmly produce the correct ones. No visual feedback of the measured signal was provided during the procedure. Fig. 2 Prototype device to capacitively detect frowning and lifting of eyebrows Fig. 3 The circuit board layout of the electrodes Med Biol Eng Comput (2010) 48:39-47 ...

Context 6

... the actual trial, the participant activated the muscles according to instructions given as synthesized speech. Each instruction stated the activation and the target intensity. Then a beep sound was played to indicate that the activation task started. A vertical bar whose height indicated the current activation intensity was shown to the participant, and the target level was highlighted as shown in Fig. 3. After the participant had held the muscle activation in the target level for 2 seconds, another beep sound was played to indicate the successful completion of the task. Participants were given 10 seconds for completing each task, and between tasks they were instructed to relax for 5 seconds. Ten repetitions of each of the three intensities of the two target muscles of that trial were performed in randomized order. This resulted in a total length of 20 minutes for the ...

Context 7

... causes a delay. Therefore distance signals without smoothing were used when classifying the facial activity. Baseline removal was carried out to the distance signals directly. Figure 3 presents the algorithm used for solving the baseline for its removal. The baseline calculation was based on a median filter. The median can perform well in the task because the signals were expected to have longer baseline sequences than the ones resulting from facial activity. The median filter applies a logic that only selects part of the samples as baseline points for the median calculation. The selection is based on a constant false alarm rate (CFAR) processor that calculates an adaptive threshold based on the noise characteristics of the processed signal [23], [24]. The distance signal was first pre-processed with a filter that implements a differentiator, a single-pole low-pass filter with a time constant of 20 ms, and a full-wave rectifier. This makes the input suitable for the CFAR processor. The current sample is used as a test sample for the processor. The implemented version of the processor uses samples before the test sample, referred to as reference samples, for calculating the threshold. The processor also leaves out samples closest to the test sample as guard samples to reduce the information overlap between the test and reference samples. Samples closer than 1 second to the test sample were considered guard samples and the following 14 seconds were considered as the reference samples. The mean of the reference samples was then calculated and multiplied by a sensitivity parameter to obtain the adaptive threshold. The sensitivity parameter was chosen to be 0.5 in this case. The threshold was then fed to a comparator with the pre-processed test sample to find out if the test sample did not exceed it. The respective samples of the input signal were included in the median calculation by the selective median filter that had a length of 15 seconds. Finally, the baseline is calculated with a 2-second moving average filter from the median filtered signal to smooth step-wise transitions in the baseline ...

Context 8

... setting up the device, the maximum EMG inten- sity levels, i.e. maximum voluntary contraction levels, were determined. The average of 5 EMG intensity peak maxima during maximum voluntary contractions was considered as the maximum level. At the same time, the experimenter verified that the muscle activations were carried out without excessive activation of other muscles than the target muscle. The target intensity levels during the actual trial were defined as percentages of the determined maximum. The low level was considered to be 20%-40%, the medium 40%-60%, and the high 60%-80%. The upper limit of the high level was selected to be less than 100% because the maximum voluntary Fig. 3. The visual feedback of the EMG activation intensity that was shown to the participant during the experiments. The different background colors show the different intensities (none, low, medium, and high), the cyan vertical bar shows the current level, and the green highlighting shows the target level of the task. contraction cannot be expected to be easily produced multiple times. Before the actual trial was started, the participant was allowed to try activating the muscles and holding the activations at the target levels in a short practice ...

Context 9

... typical image of an eye with a detected pupil and corneal reflection is seen in Fig. 3(a), and 3(b) shows how the distortions look in the worst case. In the eye camera frames, the pupils were fully visible in 85.0 ± 9.86% of the frames with no statistically significant differences between the two locations. False alarms, i.e., detected pupils and corneal reflections when the pupil is not fully visible or the frame is distorted, were detected at a percentage of 16.6 ± 10.8% with no statistical difference be- tween the locations. The main results of the performance of the prototype system during the experiments are shown in Tables I and II. Outlier removal removed 11.3 ± 3.41% of the selections contributing to point-and-click errors. This corresponds to one outlier for every 7.18 selected targets. The distributions of the point-and-click errors are shown in Fig. 4. As a comparison to the errors shown, the accuracies required to hit the targets in the experimental set-up are 0.716 • , 0.859 • , and 1.15 • for target sizes 25, 30, and 40 millimeters, respectively. Fig. 5 shows the distributions of the lengths of the sequences of frames that were not successfully tracked during the exper- Fig. 4. Distributions of the point-and-click errors: (a) at the laboratory and (b) the office prior to outlier removal. Horizontal lines represent medians, the boxes extend from lower to upper quartiles, and whiskers up to the last error not considered an outlier. There is a statistically significant difference in the medians between the two conditions with participants 4m, 7f, 8f, and 10f. iments, either due to distorted frames or a failure in detecting the pupil and the corneal ...