In the past the number of medical sensors has increased with the advent of new types of sensors that measure various physiological parameters of patient. However, cabling of these sensors is still an open issue. Wireless interfaces have the potential to replace cables but create new issues. How can one be sure that the values displayed on the bedside monitor are really from the wireless sensor attached to the patient and not from the neighboring room? Therefore, the physical location of a wireless medical sensor within the medical environment is an important aspect of context-awareness in the sense of . Context-aware medical sensors are helpful in medical applications. Bardram presents benefits of context-aware medical equipment e.g. context-aware hospital bed in . In  we suggest a wireless barometric sensor to track a reference point during surgery for correction of the invasive measured blood pressure. However, detection if a wireless sensor is accidentally removed from a specific area e.g. a operational theater is still a problem. Context-aware medical sensors solve this problem. However, medical sensors require small, easy to deploy and low-cost indoor localization. In our work we assume a scenario where a number of wireless sensor nodes with a barometric sensor are deployed in an unknown medical indoor environment. We show how wireless medical sensors equipped with barometric pressure sensors can establish a common room context and recognize spatial relationships without any infrastructure. The contributions of the paper are as follows:
We present an algorithm to detect same rooms or neighboring rooms by correlation of pressure disturbances.
We build a topology graph representing the context for indoor environment in a model for buildings.
We estimate distance between wireless nodes based on cross–correlation.
The rest of this paper is organized as follows: Section 2 introduces related work and motivates the need for a new approach. We will introduce our approach in combination with the implementation in Section 3. Preliminary results are provided in Section 4. In Section 5 we summarize our work and given an outlook to future work.
2 Related work
Several methods have been suggested in the past to create or establish context and localize wireless sensors. RF-based distance estimations include two-way ranging and receive signal strength indicator (RSSI) based measurements [2, 5, 6, 9]. Systems utilizing RSSI or other RF-based technologies require line of sight or experience difficulties with multipath propagation in indoor environment. In contrast, barometric sensors investigated here, provide measurements without line of sight requirements and are independent of other participants in the network. Usage of barometric sensors for context-detection and for altitude measurement was already investigated in the past. Bianchi et al. showed barometric pressure and triaxial accelerometry based fall detection in . Sankaran et al. focus on context-detection (with the states: IDLE, WALKING and VEHICLE) of a person  with a barometric sensor. In  it was shown that differential pressure measurement improves accuracy compared to absolute measurement with a single sensor. In  Patel et al. detected movement of humans in a heating, ventilation, and air conditioning (HVAC) system with 75-80% accuracy. Our work here shows, that wireless sensors can also take advantages of barometric disturbances for distance estimation and context-detection.
Related work shows that context-detection in terms of spatial relationship without prior knowledge of the environment is not feasible with current approaches. To the best of our knowledge, we are the first to propose pressure sensors for distance estimation and room context-detection.
3 Approach and implementation
This section introduces the decentralized approach for context-aware wireless medical sensors without prior knowledge or infrastructure. Our approach is based on barometric disturbances which occur when doors are opened ore closed. This is also valid for rooms with over-pressure e.g. operational theater, since opening of a door leads to decrease of pressure.
Figure 1 shows a barometric pressure measurement of two wireless sensor nodes over time. If a door between two rooms is opened, pressure increases in one room and decreases in the other room. At t = 1 s a door was opened which decreases barometric pressure in the room of approximately 0.2 mbar. In our experiments we found out, that barometric disturbances have an duration of ≈ 400 ms. In our approach we use the last N samples for signal processing. Time synchronization between nodes ensures simultaneous detection of barometric disturbances. First, barometric disturbances are detected by analysis of the normalized auto correlation function ACF(τ). If values of the ACF at τ ≠ 0 are larger than the given confidence intervals  a barometric disturbance is detected.
If a disturbance is detected, we perform cross–correlation to calculate the delay τdelay of samples between barometric measurements of two spatially separated nodes. Therefore, distance can be calculated by knowing τdelay, the sampling interval Tsample and the speed of sound v. However, applying a moving window with N samples enables multiple distance estimations di within one barometric disturbance. Therefore, we suggest the Pearson’s correlation coefficient ri as quality indicator of a single measurement. The correlation coefficient ri is a measure of linear relationship between series of values. The result is in the interval [−1, 1] where −1 and 1 indicate a total negative or total positive correlation, respectively. Values of |ri| close to unity indicate strong linear dependency and therefore a reliable barometric measurement. Positive values of ri indicates that sensors are placed in the same room. For the example shown in Figure 1 rmax = 0.98. Negative correlation indicates that sensors are in neighboring rooms. Distance d between two wireless sensor nodes is estimated by weighting the estimated distances di of a single barometric disturbance by following equation:(1)
where Ri stand for the normalized weights i.e. ΣRi = 1. The proposed barometric measurement is implemented at our proposed medical sensor . For our practical evaluation we selected the 24 bit MS5803-01BA pressure sensor from Measurement Specialties with a sampling interval of T = 24 ms and N = 20 samples ( 0.48 s) to fully cover a single barometric disturbance.
In our evaluation four wireless sensors are distributed in different rooms on one floor, as it might be the case in a hospital. We demonstrate context detection of wireless medical sensors without prior knowledge about environment or infrastructure. Evaluation of our approach shows that removed sensors from a specific area e.g. the operational theater can be detected. Additionally, we perform distance measurements based on barometric disturbances. Our experimental setup is shown in Figure 2.
For the first experiment two nodes are placed in neighboring rooms separated by a door. Figure 3 shows the result of relative barometric pressure measurement from time 4 s to 9 s for node 4 and node 1 for an experiment. Opening of the door happened between 4 s and 5 s and between 8 s and 9 s. The maximum change of barometric pressure caused by opening a door is −0.08 mbar and −0.1 mbar for node 4 and node 1 respectively. Barometric disturbances caused by closing the door are not visible in this figure, since we closed the door gently. The negative correlation coefficient ri = −0.88 indicates that sensors 4 and node 1 are in neighboring rooms.
Additionally, we evaluated the performance of distance estimation between node 2 and node 3. Barometric disturbances are created by closing a door on this hallway. Distances between nodes are varied between 0 m and 24 m. Figure 4 shows the barometric pressure measurement of two nodes in the same room with a distance of d = 24 m. Maximum change of barometric pressure is −0.1 mbar for both nodes. Additionally, pressure curves are displaced against each other as assumed in Section 3.
Our preliminary results are summarized in Table 1 and show that the estimated distance corresponds with the true distance d and has a maximum error of ≈ 3 m. One reason for distance errors is the limited sampling interval Tsample = 0.024 ms which leads to limited resolution within our cross–correlation.
Figure 5 shows the resulting topology graph which is automatically built with the metrics distance and correlation coefficient for the edges. A topology graph shows spatial relationships between medical sensors and specific rooms.
Context of medical sensors is estimated by the correlation coefficient, estimated distance between nodes and the detection of events like opening and closing doors. Our approach enables medical sensors in an unknown environment to detect neighboring rooms, same rooms and estimate distances between other medical sensors. This en ables a low-cost and easy to deploy solution to detect e.g. removal of sensors from an operational theater. Hence, it increases robustness and safety while having the benefits of wireless sensors. Additionally, our approach detects events like opening doors as well as open doors.
5 Conclusion and future work
In this paper we presented a novel approach for automated context-detection for wireless medical sensors in an unknown environment without infrastructure. We showed that barometric pressure disturbances caused by opening and closing doors can be measured. Our approach processes pressure signals between wireless medical sensors and enables detection of same rooms, neighboring rooms and estimation of distances between wireless medical sensors. This enables medical sensors to detect if they are accidentally removed from a specific area e.g. the operational theater. Furthermore we can create a topology graph of distributed wireless medical sensors with our approach. Preliminary results show that distance estimation has an maximum error of ≈ 3 m. In the next step, we will increase the sampling rate of barometric pressure measurement to increase resolution of distance estimation.
This publication is a result of the ongoing research within the LUMEN research group, which is funded by the German Federal Ministry of Education and Research (BMBF, FKZ 13EZ1140A/B). LUMEN is a joint research project of Lübeck University of Applied Sciences and Universität zu Lübeck and represents an own branch of the Graduate School for Computing in Medicine and Life Sciences of Universität zu Lübeck. This publication is also a result of the research work of the Center of Excellence CoSA in the project m:flo, which is funded by Bundesministerium für Wirtschaft und Energie (BMWi, FKZ KF3177201ED3).
Jakob E Bardram. Applications of context-aware computing in hospital work: examples and design principles. In Proceedings of the 2004 ACM symposium on Applied computing. ACM, 2004. Google Scholar
K. Benkic, M. Malajner, P. Planinsic, and Z. Cucej. Using RSSI value for distance estimation in wireless sensor networks based on ZigBee. In Systems, Signals and Image Processing, 2008. IWSSIP 2008. 15th International Conference on, June 2008.
Federico Bianchi, Stephen J Redmond, Michael R Narayanan, Sergio Cerutti, and Nigel H Lovell. Barometric pressure and triaxial accelerometry-based falls event detection. IEEE Transactions on Neural Systems and Rehabilitation Engineering, (6), 2010. Google Scholar
Christian Bollmeyer, Mathias Pelka, Hartmut Gehring, and Horst Hellbrück. Evaluation of Radio Based, Optical and Barometric Localization for Indoor Altitude Estimation in Medical Applications. In The Fifth International Conference on Indoor Positioning and Indoor Navigation 2014 (IPIN 2014), Busan, Korea, 2014. Google Scholar
Matteo Cypriani, Frédéric Lassabe, Philippe Canalda, and François Spies. Open wireless positioning system: A wi-fibased indoor positioning system. In Vehicular Technology Conference Fall (VTC 2009-Fall), 2009 IEEE 70th, 2009.
A Faheem, R. Virrankoski, and M. Elmusrati. Improving RSSI based distance estimation for 802.15.4 wireless sensor networks. In Wireless Information Technology and Systems (ICWITS), 2010 IEEE International Conference on, 2010.
Simon S Haykin. Adaptive filter theory. Pearson Education India, 2008. Google Scholar
Shwetak N Patel, Matthew S Reynolds, and Gregory D Abowd. Detecting human movement by differential air pressure sensing in hvac system ductwork: An exploration in infrastructure mediated sensing. In Pervasive Computing.
Christof Rohrig and M Muller. Localization of sensor nodes in a wireless sensor network using the nanoLOC TRX transceiver. In Vehicular Technology Conference, 2009. VTC Spring 2009. IEEE 69th. IEEE, 2009. Google Scholar
Kartik Sankaran, Minhui Zhu, Xiang Fa Guo, Akkihebbal L Ananda, Mun Choon Chan, and Li-Shiuan Peh. Using mobile phone barometer for low-power transportation context detection. In Proceedings of the 12th ACM Conference on Embedded Network Sensor Systems, 2014. Google Scholar
Bill N Schilit and Marvin M Theimer. Disseminating active map information to mobile hosts. Network, IEEE, 8(5), 1994. Google Scholar
About the article
Published Online: 2015-09-12
Published in Print: 2015-09-01
Conflict of interest: Authors state no conflict of interest. Material and Methods: Informed consent: Informed consent is not applicable. Ethical approval: The conducted research is not related to either human or animals use.