27 November 2012
Researchers at Philips Research Netherlands have presented a wideband human presence sensor with integrated audio and ultrasound processing. The detector shows substantive improvement in occupancy determination under silent and stationary occupancy.
Occupancy-based control of lighting, heating, ventilation and air-conditioning systems can result in significant energy savings in industrial, commercial and home environments. Presence detection is also an important part of intrusion detection and home automation systems, and localized presence sensing is useful in high quality hands-free audio communication applications - to virtually steer a microphone array towards the location of a desired talker while suppressing interference and background noise.
Occupancy detection can be achieved using a number of techniques, such as audio, ultrasound and passive infrared (PIR). Audio sensors can determine presence from audio activity but will naturally fail when an occupant is silent. Ultrasonic and PIR sensors determine presence based on movement but suffer in detection performance when the occupant is stationary. By fusing signals over the audible and ultrasonic frequency range, the team can detect silent and stationary occupancy with high accuracy.
Ultrasonic sensors are known to be more sensitive than PIR sensors, but this leads to a trade-off between the probabilities of detection and of false alarms (if the detector is too sensitive). Conversely, while large user movements are reliably detected, small movements such as when typing on a keyboard may result in missed detections, for a given false alarm probability.
Presence detection based on ultrasonic sensors is an active procedure, as opposed to techniques based on PIR or audio alone. Analogously to sonar in navigation, occupancy is determined using the reflections of a transmitted pulse, collected over consecutive transmission periods. The difference signal computed over these periods contains non-zero components (a phase shift) only in the presence of moving objects, as the signals from static objects will return in phase and cancel out. Additionally, as the processing is based on collected reflections, the occupants need not be in the line-of-sight of the ultrasound sensor, in contrast to PIR-based systems.
Audio activity techniques are also capable of detecting non-stationary audio events only. For example, in an office environment, stationary background noise such as that from air conditioning units can be ignored and only non-stationary events such as speech or typing will be detected.
The group’s detector extends the range of ultrasonic detectors into this audio region. As Sriram Srinivasan, the lead author, explained: “although the wideband sensor is primarily designed to operate at ultrasonic frequencies, for example 40 kHz, there is sufficient energy in the audible frequencies to permit audio event detection, as shown in our paper.” Therefore, by exploiting both audio and ultrasound modalities, it is possible for them to determine accurately occupancy under both silent and stationary presence. Srinivasan also said that the “relatively poor signal-to-noise ratio (SNR) at the audible frequencies compared to the SNR of a conventional microphone is acceptable as the goal is not hi-fidelity audio capture but audio-based presence detection.” Using a single sensor instead of separate audio and ultrasound sensors also simplifies product development and reduces cost, meaning the prototype may be easily brought to commercial use.
Now that the group have proved the concept with a single receiver, they are hoping to extend their detector to an array - “this will offer more flexibility and enable tasks such as direction-of-arrival estimation”, said Srinivasan. Specifically, he said that “the small size of commercially available sensors permits beamforming even at frequencies as high as 40 kHz, and the resulting localized occupancy information can be used to design sophisticated control strategies for various applications.”
The group hopes that their design represents a shift towards multi-modal and distributed sensing, where data from multiple and varied sensors are aggregated to enable complex and accurate control strategies. Looking further ahead, Srinivasan explained that “a network of such sensors should be self-configuring, not requiring an additional commissioning step, to facilitate rapid deployment.”
This article is based on the Letter: Presence detection using a wideband audio-ultrasound sensor (new window).
A PDF version (new window) of this feature article is also available.
Browse or search all papers in the latest or past issues of Electronics Letters on the IET Digital Library.