As computing continues to simultaneously increase in power, shrink in size, and decrease in cost, smart devices are making their way into nearly every aspect of our lives. But for smart devices to be very smart, they need sensors to give them information about the world around them. Of course a great many sensor options exist, but not all are appropriate for home use where people generally want sensors to blend in with the surrounding environment and not compromise their privacy.
An interesting new concept in environmental sensing has recently surfaced in a SIGCHI conference paper that may just check all the boxes for an inconspicuous in-home sensing device. Capacitivo is an interactive fabric with embedded capacitive sensors that is capable of recognizing many non-metallic objects that it comes into contact with.
The sensor-laden fabric is wired via conductive thread to a custom sensing board that contains the capacitive sensing circuits as well as an Arm-based microcontroller for data processing.
When an object comes into contact with Capacitivo, the electric field generated by the electrodes causes an electric displacement within the object. Measuring this electric displacement can reveal a signature that is distinctive to each type of object.
A random forest classifier was chosen for the object recognition component. This method was found to be accurate, robust, and much more power efficient than alternate methods (e.g. Hidden Markov Models, Convolutional Neural Networks).
The Capacitivo team demonstrated several uses for the new device. In one scenario, they lined the pocket of a jacket with the interactive fabric. The fabric was trained to recognize AirPods headphones and was able to alert the owner as to their whereabouts if they were lost.
In another demonstration, the researchers lined a fruit bowl with their capacitive fabric. Based on the fruits detected in the bowl, they were able to provide customized suggestions for fruit smoothie recipes.
In a more practical demonstration, Capacitivo was able to detect if household plants needed to be watered. Also interesting was an application in which a credit card could be set on the table next to a laptop, which would cause the credit card information to automatically populate in an online shopping form.
In its current form, Capacitivo fabrics must be trained in the exact configuration in which they will be used. For example, if the fabric is curved to fit a bowl, it will no longer be able to accurately detect the objects it was designed to recognize if that fabric is flattened out. The device is also limited to sensing only non-metallic objects.
Capacitivo is in the early stages of development, and not quite ready for general use, but it represents a new and very interesting means for devices to collect information about their environment.