Capacitive Facial Activity Measurement

Abstract

The human facial activity consists of voluntary and spontaneous behavior that can be measured to provide valuable information for several application domains. The objective of this thesis is to introduce a new, capacitive measurement method for the task. The motivation to develop a new method was to avoid some of the drawbacks that existing methods have. The existing ones that have been used to measure activity from the entire face are electromyography (EMG) that measures the electric activity of the facial muscles and image-based methods that use machine vision. EMG has drawbacks due to its requirement to attach electrodes to the face, whereas vision-based methods rely on using cameras and heavy computational processing to collect information about the facial behaviour. The presented capacitive method does not require physical contact to the face, the computational requirements of the needed signal processing are low, and it can be used in mobile applications because the measurement can be integrated to head-mounted devices.

The thesis includes several studies where prototypes were constructed, experiments carried out, signal and data processing methods applied, and results analysed. The method was first applied as a way to detect facial movements for human–computer interaction. It was integrated with a gaze tracker to point targets on a computer screen with the gaze and click with facial movements. Later, the method was extended from the detection of single facial movements to be feasible in the detection and classification of movements and expressions of the entire face. It was also applied to the measurement of the activation intensities of certain facial muscles.

The results of the thesis show that the new measurement method detects facial movements with a good performance. Pointing and clicking also performs well when the detection method is combined with gaze tracking. The classification of facial movements performs very well with the ones included in the experiments of the thesis, and the classification can be expected to work also with more complex facial expressions. Further, the measurement method’s performance in determining the intensities of facial muscle activations was good for ones that have a wide movement range. Finally, the thesis also states the limitations of the new measurement method and includes suggestions to overcome them and to develop the method further.