Investigating Low-Cost Wireless Occupancy Sensors for Beds

  • Andreas BraunEmail author
  • Martin Majewski
  • Reiner Wichert
  • Arjan Kuijper
Conference paper
Part of the Lecture Notes in Computer Science book series (LNCS, volume 9749)


Occupancy sensors are used in care applications to measure the presence of patients on beds or chairs. Sometimes it is necessary to swiftly alert help when patients try to get up, in order to prevent falls. Most systems on the market are based on pressure-mats that register changes in compression. This restricts their use to applications below soft materials. In this work we want to investigate two categories of occupancy sensors with the requirements of supporting wireless communication and a focus on low-cost of the systems. We chose capacitive proximity sensors and accelerometers that are placed below the furniture. We outline two prototype systems and methods that can be used to detect occupancy from the sensor data. Using object detection and activity recognition algorithms, we are able to distinguish the required states and communicate them to a remote system. The systems were evaluated in a study and reached a classification accuracy between 79 % and 96 % with ten users and two different beds.


Capacitive proximity sensors Bluetooth LE Smart furniture Home automation 

1 Introduction

Bed or seat occupation sensors are commonly available in intensive care facilities to give audible alerts when a patient tries to exit a chair or a bed. Caretakers are able to react on this signal can prevent falls during the getting up process. Another potential application area are home automation systems. A suitably placed occupancy sensor could be used to control heating and lighting to save energy. The most common commercial sensor is a pressure mat that is placed below the mattress [1]. These sensors are typically closed system that have no external communication method. Additionally, they may fail for light persons or stiff mattresses. They may also be difficult to switch between different beds. Modern sensing devices combine sensors, such as accelerometers and communication systems including Bluetooth in small form factors, tuned towards energy efficiency and portability. In this work we want to investigate if these systems allow us to create affordable and versatile occupancy sensors for beds or other forms of seating.

We would like to evaluate how well non-pressure sensing is suited for occupancy detection. Two factors are considered. The first is the movement of the bed frame or slatted frame below the bed when a person enters the bed. An accelerometer or motion sensor that is sufficiently sensitive can detect this movements and detect entry and exit events. The second is the usage of presence sensors that can be placed below the bed, but are still able to detect the presence of human bodies. We investigate two sensor technologies. The aforementioned accelerometer detects changes in acceleration of an object, while having a high sensitivity. We can use them to analyze entry and exit events on a bed. The second type are capacitive proximity sensors that detect the presence of a human body over a distance.

We introduce two methods that use the acquired sensor data to detect the necessary events. The accelerometer uses a threshold-based feature for activity tracking, while the capacitive sensors require initial calibration, drift compensation and several thresholds for distinguishing poses. These methods have been implemented into two prototype devices created for this project. One is based on a plain LightBlue Bean that integrates an accelerometer with free access to the data [2]. The second is an Arduino with an attached capacitive sensor based on the Capacitive Sensing library [3].

In an evaluation, we attach the devices to different objects suitable for occupancy sensing, as shown in Fig. 1. We first test if the systems are generally suited to detect occupancy on five objects. In a second evaluation we tested the classification accuracy of the occupancy sensing on two different beds with ten users. In this case we collected 200 samples of occupied and unoccupied states.
Fig. 1.

Potential scenarios for wireless occupancy systems. Bed on top left, office chair on top right, wheelchair on bottom left, and couch on bottom right.

This paper proposes the following scientific contributions:
  • An occupancy detection method, based on presence sensing for capacitive proximity sensors

  • An activity recognition method for accelerometers to identify occupancy

  • Two prototypes implementing the developed methods have been built

  • Two evaluations were performed - one for application on different types of furniture and one for classification accuracy on beds with a larger sample size.

2 Related Works

There is a large body of research that has evaluated how to create smart furniture that is able to detect the presence, posture or even physiological parameters of its occupants. Harada et al. use pressure mats to create pressure images [4, 5]. They extract a number of features from these images, in order to detect the current posture of the persons on the pressure mat. In an extension they even provide a method to reconstruct a 3D model of the human body on the pressure mat and detect motion patterns.

Hong et al. created a sensing chair that uses a pressure mat to detect postures [6]. They create pressure maps from the sensor data and calculate eigenpostures, a feature based on eigenvectors. Using training data from 20 users they achieve an accuracy between 90.3 % and 99.8 %.

The Health Chair by Griffiths et al. is an office chair equipped with an array of sensors used to detect occupancy, activities, and physiological activities on it [7]. It incorporates pressure sensors and ECG in the armrests to detect the heart rate. We have been working with sensing chairs in the past that support posture recognition or exercises on the seat [8, 9]. Additionally, we previously worked with capacitive sensors under beds that were used for a more fine-grained posture detection, the use of flexible materials for measurement, or tailored towards the recognition of sleep phases [10, 11, 12]. In this work we want to provide a simple and portable method to just track occupancy without additional features.

Capacitive proximity sensors detect the body by its influence on a generated low-intensity electric field. These sensors can detect the body over a distance and are thus suitable for installation below the bed frame. A prototype by MacLachlan was able to detect objects at a distance of 1.5 m [13]. It is a very versatile technology that can be used flexibly, by modifying materials, geometry, circuit design, and processing methods used. We give an overview of potential use cases and design considerations that are the basis for the capacitive prototype of this work [14].

The advent of smartphones has spawned numerous applications that use the installed accelerometers or microphones to detect motion during sleep [15, 16]. The phone is usually placed somewhere on the mattress, e.g. below the pillow and tracks movement either by vibrations of the mattress or sounds generated by the user.

Bluetooth Low Energy or BLE is a communication standard for low-power devices that need to communicate over medium distances of up to 30 m with other systems. It is commonly used for location tags or iBeacons that can be used to support mobile devices in indoor navigation [17]. Recently it has been used to provide small programmable microcontrollers with low-power communication facilities. One example device is the LightBlue Bean, an Arduino-compatible microcontroller without wired connections that is powered by a coin cell [2].

3 Presence Sensing Using a Capacitive Proximity Sensor

In general we can distinguish between the calibration phase of the occupancy sensor that is performed initially and the execution phase of the sensor performed during normal operation. Presence sensing with a capacitive proximity sensor requires four different processing steps, as shown in Fig. 2.
Fig. 2.

Data processing of presence sensing using a capacitive proximity sensor

Environment calibration is necessary, as the basic capacitance measurement of a sensor is based on the environmental parameters it resides in, as well as various other factors, such as electrode material or duration of measurement [14]. Usually this does not require more than taking a certain amount of samples and calculating the initial average.

Occupancy calibration is performed on the first object detection. After the sensor is installed under a surface and calibrated towards the environment, a person is sitting or lying on the object. The resulting sensor value is indicative of the object being occupied and stored for later use. Here we can collect a number of samples and calculate an average.

Drift compensation is the process to account for changes in capacitance caused by changes in the environment. E.g. if the temperature of the system increases or the humidity in the room is lowered, the resulting sensor values will change. Drift compensation is a process that analyses the sensor values over a longer time and applies changes to the initial environment calibration. This is crucial for applications, where the sensor is turned on for a long time.

Occupancy sensing is based on a simple threshold method. Typically, a threshold somewhere between the value after environment calibration and the occupancy values is calculated. In this case the threshold is put at 50 % of the occupancy value. If the sensor values exceed this threshold, occupancy is detected. Additionally, we specify that the threshold has to be exceeded for a specific number of samples or a time frame. In our case the system is considered occupied if the occupancy threshold is exceeded for at least one second.

4 Activity Recognition Using Accelerometers

Data processing to recognize activities from accelerometers is simpler compared to capacitive sensing. The sensors are available in packages that already perform all calibration routines internally. We receive angular values that depict the acceleration of the sensor in x, y, z direction. Moving the sensor will cause these values to change. We define motion as the difference in subsequent acceleration readings. In our application the direction of movement is not very important, as the system is hanging freely. For each sample we calculate the sum of values to get the overall acceleration. The difference of these sums over three samples is used to calculate the velocity of the motion (Fig. 3).
Fig. 3.

Data processing of activity recognition using accelerometers

Enter/Leave calibration is performed the first time a person is entering the object (or sitting on the object). The recorded velocity is stored. In our system we use the maximum velocity recorded during several sit/lie (get up/stand up) events.

Enter/Leave detection now looks for motion velocities that result in similar values to the calibration phase. Since we want to account for variability the threshold to be exceeded is put at 80 % of the maximum velocity of the calibration phase.

4.1 Prototype Systems

Two prototypes have been created for this work. The accelerometer activity system was implemented on the LightBlue Bean, an Arduino-compatible microcontroller tailored for Bluetooth LE communication [2]. We attached wires to the system, that can be used to connect external systems or in our case, to attach the prototype to the piece of furniture where we want to register occupancy. The accelerometer is integrated on the board, as well as a Bluetooth chip. It is powered using a coin cell battery. Figure 4 gives an overview of the LightBlue Bean, the attached wires and some important components. The device is programmed via Bluetooth, usually from a smartphone. We use the integrated accelerometer and an activity recognition method based on a single-direction movement of the sensor.
Fig. 4.

Close up picture of LightBlue Bean and the hanger

Fig. 5.

Arduino with resistors and printed circuit board installed on a breadboard

For the second system the Arduino Capacitive Sensing library was used to create a single channel capacitive sensor [3]. A dual layer PCB is used as electrode and connected to a digital input using a large 40 MOhm resistor (Fig. 5). This enables sensing at distances of up to 40 cm. This capacitive system distinguishes three states that have to be initially calibrated - “sitting”, “lying” and “not on bed”.

The cost of both systems is comparatively low. The LightBlue Bean retails for $30, while the components of the capacitive system together cost around $50. It has to be noted that the latter could be reduced significantly if an integrated system would be used.

5 Evaluation

In our evaluation we test how well both systems recognize entering and leaving events for different persons and different beds. The evaluation was two-fold. At first the system was tested with five different pieces of furniture by two different users.
  1. 1.

    Office chair - is an office chair with backrest and armrests that has a gas spring. Sensors are attached at the bottom of the seat.

  2. 2.

    Wooden chair - is a thinly cushioned chair with wooden seat and backrest and metal legs. Sensors are attached at the bottom of the seat.

  3. 3.

    Wheel chair - is a basic wheel chair with leather seat area. Sensors are glued to leather seat.

  4. 4.

    Bed #1 - has a thin mattress and a flexible slatted frame. Sensors are attached on slatted frame.

  5. 5.

    Bed #2 - has a thick mattress on a solid slatted frame. Sensors are attach on slatted frame.

Each took a seat five times, stayed for ten seconds and got up again, leading to an overall number of 20 recognizable events. In the second part ten different users occupy the two different beds. Here everybody takes a seat ten times, waits ten seconds and gets up again. The results of both evaluations are shown in Table 1.
Table 1.

Results of evaluations one and two with five different objects and 2–10 users


No. samples

Recall accelerometer

Recall capacitive

Office chair


90 %

85 %

Wooden chair


55 %

100 %

Wheel chair


85 %

85 %

Bed #1


75 %

95 %

Bed #2


50 %

90 %

Bed #1 (10 users)


91 %

96 %

Bed #2 (10 users)


79 %

93 %

The accelerometer system is able to detect entry and leaving events with a good success rate - however struggles at differentiating both, if one of the events results in an increased vibration. Overall, the detection rate is not as good, as for capacitive sensors, with the exception of the office chair. Here, the distance between the bottom of the chair and the seat is somewhat far, making it more difficult to detect for the capacitive sensors, while the accelerometer system benefits from the gas spring that results in a movement of the whole seat area. The accelerometer struggles when the structure is very rigid or the person is careful when taking a seat, as this results in reduced movement of the object. The capacitive senor does not have this limitation, but can be disturbed by metal parts near its location, as was the case for the office chair and the wheel chair.

6 Conclusion and Future Work

On the previous pages we have introduced two low cost systems for occupancy detection on objects, particularly beds. One system uses an accelerometer for detecting movement, while the second uses a capacitive proximity sensor to detect objects in range. Both systems have advantages and disadvantages, with the accelerometer system being cheaper in the current prototyping stage, while the capacitive sensing system was more reliable, with detection rates between 85 % and 100 % in the small sample evaluation and 93 % to 96 % in the larger second evaluation.

In the future we would like to use more precise capacitive systems that are better integrated to combine high accuracy with low cost. OpenCapSense by Grosse-Puppendahl et al. is a proximity sensing toolkit with high precision that might be suitable [18].


  1. 1.
    Tunstall: Tunstall Bed Occupancy Sensor. Accessed 26 Feb 2016
  2. 2.
    Punch Through Design: The LightBlue Bean. Accessed 26 Feb 2016
  3. 3.
    Badger, P.: Capacitive Sensing Library.
  4. 4.
    Harada, T., Mori, T., Nishida, Y., Yoshimi, T., Sato, T.: Body parts positions and posture estimation system based on pressure distribution image. In: Proceedings 1999 IEEE International Conference on Robotics and Automation (Cat. No. 99CH36288C), pp. 968–975. IEEE (1999)Google Scholar
  5. 5.
    Harada, T., Sato, T., Mori, T.: Human motion tracking system based on skeleton and surface integration model using pressure sensors distribution bed. In: Proceedings Workshop on Human Motion, pp. 99–106 (2000)Google Scholar
  6. 6.
    Tan, H.Z., Slivovsky, L.A., Pentland, A., Member, S.: A sensing chair using pressure distribution sensors. IEEE/ASME Trans. Mechatron. 6, 261–268 (2001)CrossRefGoogle Scholar
  7. 7.
    Griffiths, E., Saponas, T.S., Brush, A.J.B.: Health chair: implicitly sensing heart and respiratory rate. In: Proceedings of the 2014 ACM International Joint Conference on Pervasive and Ubiquitous Computing - UbiComp 2014 Adjunct, pp. 661–671. ACM Press, New York (2014)Google Scholar
  8. 8.
    Braun, A., Frank, S., Wichert, R.: The capacitive chair. In: Streitz, N., Markopoulos, P. (eds.) DAPI 2015. LNCS, vol. 9189, pp. 397–407. Springer, Heidelberg (2015)CrossRefGoogle Scholar
  9. 9.
    Braun, A., Schembri, I., Frank, S.: ExerSeat - sensor-supported exercise system for ergonomic microbreaks. In: De Ruyter, B., et al. (eds.) AmI 2015. LNCS, vol. 9425, pp. 236–251. Springer, Heidelberg (2015). doi: 10.1007/978-3-319-26005-1_16 CrossRefGoogle Scholar
  10. 10.
    Djakow, M., Braun, A., Marinc, A.: MoviBed - sleep analysis using capacitive sensors. In: Antona, M., Stephanidis, C. (eds.) UAHCI 2014, Part IV. LNCS, vol. 8516, pp. 171–181. Springer, Heidelberg (2014)Google Scholar
  11. 11.
    Braun, A., Heggen, H.: Context recognition using capacitive sensor arrays in beds. In: Proceedings AAL-Kongress (2012)Google Scholar
  12. 12.
    Rus, S., Grosse-Puppendahl, T., Kuijper, A.: Recognition of bed postures using mutual capacitance sensing. In: Aarts, E., et al. (eds.) AmI 2014. LNCS, vol. 8850, pp. 51–66. Springer, Heidelberg (2014)Google Scholar
  13. 13.
    MacLachlan, R.: Spread Spectrum Capacitive Proximity Sensor. Accessed 26 Feb 2016
  14. 14.
    Braun, A., Wichert, R., Kuijper, A., Fellner, D.W.: Capacitive proximity sensing in smart environments. J. Ambient Intell. Smart Environ. 7, 1–28 (2015)Google Scholar
  15. 15.
    Jones, C., Campbell, S., Zone, S.: Familial advanced sleep-phase syndrome: a short-period circadian rhythm variant in humans. Nat. Med. 5(9), 1062–1065 (1999)CrossRefGoogle Scholar
  16. 16.
    Krejcar, O., Jirka, J., Janckulik, D.: Use of mobile phones as intelligent sensors for sound input analysis and sleep state detection. Sensors 11, 6037–6055 (2011)CrossRefGoogle Scholar
  17. 17.
    Apple Inc.: iBeacon for developers. Accessed 26 Feb 2016
  18. 18.
    Grosse-Puppendahl, T., Berghoefer, Y., Braun, A., Wimmer, R., Kuijper, A.: OpenCapSense: a rapid prototyping toolkit for pervasive interaction using capacitive sensing. In: 2013 IEEE International Conference on Pervasive Computing and Communications, PerCom 2013, pp. 152–159 (2013)Google Scholar

Copyright information

© Springer International Publishing Switzerland 2016

Authors and Affiliations

  • Andreas Braun
    • 1
    • 2
    Email author
  • Martin Majewski
    • 1
  • Reiner Wichert
    • 1
  • Arjan Kuijper
    • 1
    • 2
  1. 1.Fraunhofer Institute for Computer Graphics Research IGDDarmstadtGermany
  2. 2.Technische Universität DarmstadtDarmstadtGermany

Personalised recommendations