Just Face It

Do you remember when hotel rooms used to have actual, physical keys for unlocking the door? Keycards and other keyless entry systems have long since taken over as the standard access control method. So much so that if you were handed a key during a hotel check-in, you might start to wonder if the travel site you made the booking on led you astray.

There are a lot of good reasons for this change. Security, for one, is much better when the key codes can be rotated after each guest’s stay. Convenience is another major benefit — fumbling around with a key when your hands are loaded down with luggage or groceries is no fun at all, to say the least.

So then, with advantages such as these, why do we still predominantly use physical keys to access our homes? Keyless entry systems are rising in popularity in residential settings, but they are still deployed in a small minority of homes. Perhaps people do not trust them? Or maybe they think that they are too complicated? Whatever the case may be, engineer Thomas Vikström has developed a custom keyless entry system that may change the minds of many naysayers. Rather than relying on a keycard or access code, Vikström’s device uses facial recognition to automatically unlock the door as an authorized individual steps up to it.

At the heart of the access system is a Particle M.2 SoM Evaluation Board equipped with a B524 B-SoM module for cellular and Bluetooth connectivity. A tiny and inexpensive Person Sensor from Useful Sensors was also included to simplify the facial recognition aspect of the project. And to handle the mechanical aspect of unlocking a door, Vikström leveraged a servo-powered linear actuator kit.

Vikström’s game plan involved designing a device that can continually scan for faces. When a face is recognized, the system will then compare it with a set of up to eight known faces to see if there is a match. If a match is found, the device will send a signal to the linear actuator to unlock the door, and it will simultaneously send a mobile notification to the homeowner, indicating the identity of the person that has accessed their home.

With just a few components handling the functions of the device, the wiring was very simple. After assembly, the Person Sensor was put through a calibration process, which allows it to remember the faces of up to eight individuals. After that process, the Person Sensor is capable of reporting the presence of known individuals' faces to the M.2 SoM Evaluation Board via an I2C connection. The Particle board then triggers the linear actuator to unlock the door and, via a Particle webhook, sends a smartphone notification through Pushover.

Vikström has presented a very interesting project using components that we can individually vet so that we feel safe with the security of the system and know it is not sending our data off to some unknown cloud-based server farm. However, it was noted that the Person Sensor is not really intended for security-critical applications, so you may want to swap that out for a more powerful computer vision system before actually deploying it in your home. A 3D-printer would also come in handy to provide for some weatherproofing of the device.

If this is a platform you would like to enhance for your own purposes, the best way to start is by reading the project write-up.