Ambient Diagnostics

People can usually sense troubles in a car from noises, vibrations, or smells. An experienced driver can even tell where the problem is. We call this kind of skill ‘Ambient Diagnostics’. Ambient Diagnostics is an emerging field that is aimed at detecting abnormities from seemly disconnected ambient data that we take for granted. For example, the human body is a rich ambient data source: temperature, pulses, gestures, sound, forces, moisture, et al. Also, many electronic devices provide pervasive ambient data streams, such as mobile phones, surveillance cameras, satellite images, personal data assistants, wireless networks and so on. The peripheral vision of the redundant information enables Ambient Diagnostics. For example, a mobile phone can also be a diagnostic tool. As the sounds generated by breathing in asthma patients are widely accepted as an indicator of disease activity [1, 2], researchers have investigated the use of a mobile phone and electronic signal transfer by e-mail and voice mail to study tracheal breath sounds in individuals with normal lung function and patients with asthma [3]. The results suggest that mobile phone recordings clearly discriminate tracheal breath sounds in asthma patients and could be a non-invasive method of monitoring airway diseases. It is challenging to extract just one bit of diagnosis (positive or negative) from massive ambient data. First, we need pivotal heuristics or domain knowledge. In many cases, the heuristics just serve as an early warning rather than an accurate examination. For example, medical studies show that snoring may be related to hypertension, cardiac dysfunction, angina pectoris and cerebral infarction. The immediate rise in systemic blood pressure during snoring has been confirmed by polygraphic recordings [60]. A ‘snoremeter’ could be added into a mobile phone because it already has a microphone inside. It would provide valuable early warnings for related diseases. Second, we need physical heuristics that effectively filter out the trivial data while only keeping the abnormities. Knowing the physical properties of the targeted system would greatly benefit a diagnosis. For example, if we know that the needle in a hay stack is metal, then we can work around the metal properties and make the hay disappear. Ideally, physical heuristics map the data to a feature space that only displays limited interesting features. Determining which