Respiratory is one of the vital signs used to monitor the progression of the illness that are important for clinical and health care fields. From home rehabilitation to intensive care monitoring, the rate of respiration must be constantly monitored as it offers a proactive approach for early detection of patient deterioration that can be used to trigger therapeutic procedures alarms. The use of invasive procedures based on contact transducers is typically necessary to measure the quantity. Nevertheless, these procedures might be troublesome due to the inconvenience and sensitivity of physical contact. Therefore, non-contact human breathing monitoring as a non-invasive procedure is important in long term intensive-care and home healthcare applications. In this paper, respiratory signals from two type of resting activities had been acquired and proposed a Deep Neural Network (DNN) model that can classify the respiratory signal into inhalation and exhalation signal. Several pre-processing techniques has been done onto the signal before it is implemented into the proposed model. The average recognition rate of the respiratory signal using the proposed method was 84.1% when the subject was sleeping and 83.8% when awake.
