Abstract— Quality of sleep is an important index of wellbeing
and health. Irregular sleep patterns are often associated with stress
and disorders such as cardiovascular disease, diabetes, depression,
sleep apnea and obesity. In addition to key physiological indices,
body movements and posture during sleep are also important
for assessing causal relationship of irregular sleep patterns and
underlying health issues. In this paper, we explore the feasibility
of using a single accelerometer strapped onto the chest to detect
posture and cardio-respiratory parameters during sleep. An efficient
movement detector suitablefor on-node implementation is developed
to distinguish static postures from dynamics movements. When in
static postures, a linear discriminant analysis (IDA) classifier is
used to further divide the static postures into four common sleeping
positions. Simultaneously, both heart rate and respiratory rate are
extracted from the acceleration signal. A small cohort of 7 healthy
subjects were recruited for lab-controlled experiments to evaluate
the performance of our proposed methods. ECG signal and K4b2
system's V02 measurements were also collected to extract heart
rate and respiratory rate as the ground truth for comparison. An
overall classification accuracy of 99% is achieved for recognising
the correct sleeping positions. Good matches to ground truths were
also obtained for the derived cardiac and respiratory rates.