Jing Zhai

Signal processing quantification of changes in the blood volume pulse (BVP) waveform due to exercise

Observations made in the past by our group confirmed that the blood volume pulse (BVP) waveform recorded using an infrared finger photoplethysmograph (PPG) undergoes changes as the subject performs physical exercise. In particular, the Dicrotic Notch of the BVP waveform has been observed to become less prominent in connection with the performance of exercise. There is an interest in characterizing those changes through a single parameter to measure the level of exercise the subject has reached, at any time during an exercise session. This paper reports on the comparison of three Digital Signal Processing approaches designed to reflect the BVP waveform changes through a single parameter, which could be obtained automatically from the digitized BVP signal. For this study BVP measurements were taken from 10 subjects, as they engaged in upper-body exercises and rested afterwards.

Significance of Pupil Diameter Measurements for the Assessment of Affective State in Computer Users

The need to provide computers with the ability to distinguish the affective state of their users is a major requirement for the practical implementation of Affective Computing concepts. The determination of the affective state of a computer user from the measurement of some of his/her physiological signals is a promising avenue towards that goal. In addition to the monitoring of signals typically analyzed for affective assessment, such as the Galvanic Skin Response (GSR) and the Blood Volume Pulse (BVP), other physiological variables, such as the Pupil Diameter (PD) may be able to provide a way to assess the affective state of a computer user, in real-time. This paper studies the significance of pupil diameter measurements towards differentiating two affective states (stressed vs. relaxed) in computer users performing tasks designed to elicit those states in a predictable sequence. Specifically, the paper compares the discriminating power exhibited by the pupil diameter measurement to those of other single-index detectors derived from simultaneously acquired signals, in terms of their Receiver Operating Characteristic (ROC) curves.

Exploring face recognition by combining 3D profiles and contours

Most of the face recognition research performed in the past used 2D intensity images obtained by a photographic camera as the data format for processing, but the algorithms developed based on 2D images are not robust to changes of the conditions in which the images are taken, like the illumination of the environment and the orientation of the subject. With the development of 3D imaging techniques, 3D face recognition is becoming a natural choice to overcome the shortcomings of 2D face recognition, since a 3D face image records the exact geometry of the subject, invariant to illumination and the orientation changes. In this paper, a new algorithm for automatic face recognition, based on the characterization of faces by their contours and profiles, is proposed. Experiments show that the central vertical profile and the contour are both very useful features for face recognition. When combined, better recognition rates can be obtained than just using any of them alone. The performance of the algorithm is also compared with that of the traditional principal component analysis method using a database of 80 subjects. Results show that our method, which characterizes a face through its central vertical profile and contour, can achieve better results and requires less computational power in processing this test database.

New classification algorithm for electromyography-based computer cursor control system

At present, a three-input electromyography (EMG) system has been created to provide real-time, hands-free cursor control. The system uses the real-time spectral analysis of three EMG signals to produce the following five cursor actions: i) LEFT, ii) RIGHT, iii) UP, iv) DOWN, v) LEFT-CLICK. The three EMG signals are obtained from two surface electrodes placed on the left and right temples of the head and one electrode placed in the forehead region. The present system for translating EMG activity into cursor actions does not always discriminate between up and down EMG activity efficiently. To resolve this problem it was proposed that the three-electrode system be converted into a four-electrode system, using two electrodes in the forehead of the user, instead of one. This paper compares the effectiveness of the four-electrode system to that of the three-electrode system in classifying EMG activity into cursor actions through the use of Matlab simulations. It will be shown that the new four-electrode system produces significant improvements in classification performance.