Aspinity unveils analogML algorithms for vehicle security

Aspinity has introduced new hardware and software solutions for automotive security applications, based on its AML100 analog machine learning processor. The new solutions include a suite of analogML algorithms that target parked vehicle monitoring and a new dashcam evaluation kit that detects and records security events for weeks or more without impacting the vehicle’s battery or requiring an external power source.

(Source: Aspinity)

The new solutions target the challenge of providing always-on sensing for data collection and processing when the vehicle is parked without depleting the battery. In typical always-on security solutions, they either are highly accurate or deliver an extended parking mode but not at the same time, Aspinity said.

With the AML100 near-zero-power analogML processor, vehicles can be continuously monitored for an extended period without worrying about the vehicle’s battery, the company added.

Aspinity said its automotive evaluation kit detects relevant events more accurately than today’s commonly used g-sensor based solution.

The company recently demoed a side-by-side comparison of event detection accuracy using a dashcam with a standard g-sensor and a dashcam with a single microphone and an AML100. The solution is reported to outperform the g-sensor in detecting relevant vehicle events such as the jiggling of the door handle, a neighboring car door opening into the vehicle and a shopping cart hitting the side of the car.

The Aspinity solution uses an acoustic-only trigger with analogML algorithms that have been trained specifically to identify automotive security events. They classify events that effect the vehicle like collisions or window glass breaks, while ignoring events unrelated to the vehicle such as a car alarm, a blaring horn or a large truck driving by that can often trigger a g-sensor.

The dashcam evaluation kit uses Aspinity’s AML100-REF-1 wireless, battery-operated evaluation module that allows for fast deployment and evaluation in the cabin of a vehicle.  The module is comprised of an on-board MEMS analog microphone with connection for additional sensors, the AML100 processor, dashcam event detection software and a battery.

The kit consumes <50 µA always-on and eliminates the video recording of false events that waste power, Aspinity said. It can be used as the front end of a third-party dashcam or in-cabin OEM integrated solution that triggers a camera or sends alerts when a security event is detected.

Aspinity’s library of automotive surveillance algorithms leverages the sensor fusion capabilities of the AML100 by using combinations of analog input signals from a variety of sensors, including acoustic, piezoelectric and radar. Sensors can be selected based on the specific monitoring application, sensitivity required, and whether the sensors will be located in the cabin of the vehicle or integrated within the vehicle’s panels.

The dashcam evaluation kit is available now at Aspinity. It is available as an acoustic-only solution using the on-board microphone or as sensor fusion solution using additional external sensors.