Infineon Hopes to Pack a Punch With Battery-powered Sensor Fusion Alarm Design
By marrying a MEMS microphone, a pressure sensor, and an AI/ML algorithm, a PSoC MCU-based reference design lets engineers build very low-power smart alarm systems.
This week the Sensors Converge 2022 show is humming with technologies on displays from both sensor companies and companies with a stake in sensor system design. Sensors Converge runs June 27-30 in San Jose.
Yesterday at the show, Infineon Technologies unveiled a reference design for its battery-powered smart alarm system (SAS). According to the company, the platform blends high accuracy and very low-power consumption by relying on sensor fusion based on artificial intelligence (AI) and machine learning (ML).
The smart alarm system reference design combines a MEMS microphone, a digital pressure sensor, and a PSoC 62 MCU. Top view (left) and bottom view (right). Image used courtesy of Infineon Technologies
The company claims it is the "industry's first battery-powered AI/ML-based acoustic event detection and sensor fusion alarm system."
The reference design combines these Infineon chips:
- XENSIV IM73A135V01: a high-SNR (signal-to-noise ratio) MEMS microphone
- XENSIV DPS310: a digital pressure sensor
- PSoC 62: a microcontroller (MCU)
Aside from that hardware, the SAS is comprised of a sensor fusion software algorithm, which Infineon calls acoustic event detection (AED) technology running on the PSoC MCU. The algorithm is based on what Infineon says is precisely trained AI/ML that combines acoustic and pressure sensor data to accurately distinguish between sharp sounds inside a home and distinctive audio/pressure events.
The SAS is a sensor fusion system that combines the functions of the XENSIV IM69D130 microphone and the DPS310 barometric pressure sensor. Image used courtesy of Infineon Technologies
The company says that the AI/ML sensor fusion algorithm can also rule out many other background pressure events and sounds that could lead to false positives.
AL/ML Algorithm Understands Events
The AED technology at the heart of Infineon's SAS is sophisticated machine learning Tiny AI algorithms that manage different events.
According to the company, these events can be microphone-triggered acoustic events or pressure events triggered by the pressure sensor. In addition, combined events can be triggered by a sensor fusion architecture using input from both sensors.
This sensor fusion uses Infineon's patent-pending scheme for marrying audio with pressure sensor data so that the SAS only reacts to the right events. In other words, in sensor fusion mode, the two sensors must be triggered simultaneously for the alarm to trigger.
SAS Demo at Sensors Converge
At this year's Sensors Converge show, the SAS was among the technologies showcased at Infineon's booth. There, we spoke with Tanja Hofner, Lead Principal Hardware Engineer (IoT and Sensors) at Infineon Technologies. Hofner crafted the demo of the SAS displayed at the Infineon booth.
Importantly, the system is also more robust than conventional security systems because it can't be triggered by normal household sounds or background noise—sounds for a movie playing on TV, for instance. "The algorithm knows the difference between the sounds of drinking glass breaking and a window breaking," says Hofner.
All About Circuits Editor-in-Chief Jeff Child talking about Infineon’s SAS with Tanja Hofner, Lead Principal Hardware Engineer (IoT and Sensors) at Infineon Technologies
Hofner says it's typical for a building alarm system to use multiple layers of defense such as PIR (passive infrared) sensors or LiDAR-based sensors. For example, designers would use such technologies in this application to detect intruders before they approach a window.
Those technologies are often needed because properly detecting a window-glass breaking event is difficult. In contrast, the SAS achieves a reliable detection capability that lets you dispense with technologies like PIR and LiDAR.
Three Types of Events
According to Infineon, the SAS supports three events. The first is the "glass break" event. When a window or door glass is broken, the AED algorithm extracts the intensities in the frequency spectrum. It simultaneously captures the signature pattern of the change in pressure within the room. The device's sensor fusion software synchronizes the outputs and then triggers a glass break alarm.
The next event, called "Intruder," occurs when all windows and doors are closed, and there is a stable pressure inside the room. Once a door or window is open or closed, that causes a pressure change within the room. Since each pressure event has a signature pattern, opening or closing a door or window triggers an intruder alarm.
The third event has to do with smoke detectors and carbon monoxide detectors. Designers can use the SAS to notify users when those alarms are activated when nobody is in a building or house to hear them.
Battery Life for Up to Five Years?
The low power advantages of the SAS are thanks to its ability to be placed in a low power saving mode, only activated when key acoustic event triggers such as the sound of glass breaking occur. Meanwhile, the PSoC6 MCU's deep sleep operation also contributes to longer battery life for the alarm system. Battery life up to 5 years is possible, depending on battery configuration, says the company.
Infineon says its home security alarm SAS reference design is available today, while the board itself will be available in September 2022.