Analysis of Psychophysiological Responses Using Heart Rate Variability: Towards Real-Time Affect Recognition
View/ Open
Metadata
Show full item recordAbstract
Recent technological developments have provided innovative means for promoting health and well-being through physiological response monitoring. Heart rate variability (HRV) has arisen as a promising physiological indicator of mental health. This research contributes towards these efforts by investigating the short-term effects of increased HRV using a biofeedback exercise (paced breathing) on affective states and physiological measures to facilitate the development of real-time affect recognition systems. To enable the analysis of high-quality HRV data in real-time applications, the first study examined the reliability of automatic filtering techniques using an open-source implementation. The outcomes of this study provided a flexible control for HRV signal filtering parameters and served as the basis for the analyses in the following studies. Subsequently, the second study investigated the minimum reliable window for HRV signals based on the conditions under which the data were collected. The findings suggested that HRV measures can be analysed in segments of less than 5 minutes in all conditions. Additionally, the minimum segment differed in paced breathing compared to resting and stress. Given the physiological influence of paced breathing, the third study examined the short-term effects of a heart rate variability biofeedback (HRVB) intervention on a range of affective states (e.g., relaxation, stress), working memory, and physiological data. The findings showed a significant improvement in working memory and relaxation levels following the intervention. The last study leveraged the major findings of the previous two studies to develop robust predictive models that identified stress using supervised learning algorithms. Overall, this research demonstrates that a single HRV biofeedback session mediates physiological responses and that this mediation can be measured across a range of affective states. Moreover, it shows that stress levels can be robustly recognised using supervised learning algorithms. This research also lays the groundwork for the potential employment of HRV in real-time applications to predict affective states.
Authors
Bahameish, MCollections
- Theses [4201]