How can a 3-axis acceleration sensor detect a fall?
There are many applications in electronics that promise to make our lives more comfortable and safer. For senior citizens, a body-worn home emergency call device is an assurance that enables them to lead a good quality independent life even at an advanced age. One important feature is fall detection. With a suitable sensor, this function can be very easily integrated into a corresponding device.
MEMS (Micro-Electro-Mechanical Systems) motion sensors are ubiquitous – from smartphones to automotive safety systems – and their use within the systems into which they are integrated is just as diverse. Signals generated by the MEMS must be made interpretable and usable by circuits.
How a fall detection system works
Fall detection sensors in home emergency call systems are designed to automatically detect when a person falls and remains motionless. The background: for the sake of argument, it is assumed that emergency caller cannot make a manually triggered emergency call and that the emergency call must be made automatically. How can a 3-axis acceleration sensor detect a fall? In physical terms, a fall means that a brief moment of weightlessness occurs, i.e., an acceleration of approximately 0 g is measured, followed by a high negative acceleration (rapid deceleration). If no changes in the acceleration values are then measured for a certain time, the motionlessness of the person wearing the sensor can be noticed.
For the development of a fall detection system, it means that the impulses generated by a MEMS must be interpreted correctly. Another challenge is that very accurate values must be derived. Therefore, disturbing influences such as temperature fluctuations must be compensated. In all, a great deal of effort must be expended before the evaluation function – the algorithm for fall detection – can be developed. Luckily, modern sensor modules support the developer in this context.
Würth Elektronik follows the modern principle of ‘smart’ components. Here, part of the ‘intelligence’ is embedded in the sensor, i.e., integrated. The 3-axis acceleration sensor WSEN-ITDS, which measures 2.0 × 2.0 × 0.7 mm in an LGA package, uses a MEMS-based, capacitive measuring principle. Through an integrated temperature sensor, the sensor outputs already compensated and calibrated data. Four measuring ranges are available: ±2 g, ±4 g, ±8 g or ±16 g. With necessary register settings, the application-specific functionalities of free-fall, wake-up, activity, movement and orientation detection can be selected. The state of the specific functionalities are known via two flexible interrupt pins. This processing and provision of the measured values, which is geared towards use cases, represents a simplification for application development.
This is also the case with fall detection. Due to the availability of preset functions and selectable parameters, fall detection is effectively implemented. The advantage here is that the built-in functions minimize the continuously unnecessary retrieve of the acceleration data from the sensor to perform complex calculations.
The 3-axis acceleration sensor WSEN-ITDS has two programmable interrupt pins INT_0 and INT_1. The pins can be activated or deactivated individually. The interrupt signals from the sensor functions are routed to these two pins.
As indicated in red on the diagram (Figure 1), it is the three functions “Free Fall”, “Wake up” and “Stationary/Motion” that are needed for detecting a fall. The combination of these sensor functions delivers reliable information about whether a person is falling and whether or not they can move afterwards. The interrupt signals from these events routed to the pins INT_0 and INT_1 tell the monitoring system when to trigger the alarm.
Setting limit values
During a free-fall event, the acceleration value of all three axes moves towards zero. In the register of the sensor module, two application-specific values must be defined, at the occurrence of which the interrupt for reporting a fall is generated: the threshold value, from when a free-fall is assumed, and the duration of the free-fall. For the sensor data to be interpreted in response to the question “fall or no fall?”, the waking up of the sensor must also be defined, again by setting a threshold value of changed acceleration and a minimum duration of this impulse. To exclude the gravity vector and very low frequency noise during the detection of the wake-up event, a high pass filter output should be used.
The integrated “Stationary/Motion” algorithm is the function that informs the alarm system whether the wearer of the sensor remains motionless after the fall or is still capable of acting. The Stationary function combines the two detection events: “sleeping” and “waking up”, to register a motionlessness event. There is no separate sleep interrupt signal. It is realized by monitoring the sleep and wake-up interrupt signals. The wake-up threshold and duration parameters can be defined according to user application. When a specified number of output acceleration values from the X-, Y- and Z-axes are less than the wake-up threshold and the output values remain within this range for a specified duration, the sleep interrupt signal is generated. If a single data value from one of the axes is higher than the defined wake-up threshold and the data remains in this range for a certain duration, the wake-up interrupt signal is generated. Figure 2 illustrates how the interrupts are generated that control a fall detection application.
The four phases are as follows:
1. Before the fall, the vector sum of the acceleration values from all three axes will be close to 1 g. By also monitoring the orientation of the acceleration before and after the fall, additional information about the human fall event can be obtained.
2. In the free fall condition, the phenomenon of weightlessness always occurs at the beginning of a fall. With suitable free fall duration and threshold values, the fall of a person can be detected via the interrupt signal FF_IA. During the free fall, the acceleration tends towards the zero g level, but immediately after the free fall, a strong acceleration peak occurs due to the impact of the person hitting the ground.
3. Immediately after the fall, the person will try to move. If the fall was too severe, the person may not be able to move immediately after the fall. The duration of this event can be configured using the Stationary Detection function.
4. If the person is unable to move after a certain time (configured in the Stationary function) because they are unconscious, the interrupt signal SLEEP_STATE_IA and SLEEP_CHANGE_IA are generated. By comparing the orientation of the sensor acceleration before and after the fall, the fall detection system can be instructed to generate an alarm. If the person has moved within a certain time, the wake-up signal WU_IA is generated instead of the interrupt signal SLEEP_STATE_IA and SLEEP_CHANGE_IA. In this case, the fall detection system does not generate an automatic alarm.
As the explanations have shown, a sensor module already relieves the developers with number of tasks. The electronics and software of a fall detection system can therefore be made simpler. The power-saving sensor module, which wakes itself up in the event of defined acceleration events, also has a positive impact on the energy consumption and service life of a battery-powered system.
For more detailed information, click here.
This article was originally published on EE Times Europe.