Today kicked off a week of demo meetings at CES showcasing how SensiMLTM Analytics Toolkit can make quick work of algorithm development for IoT sensing wearable devices. Our demo suite showed first-hand the data collection, AutoML code generation, and testing of two different wearable applications that were built in under a week.
Demo #1 – Boxing Heavy Bag Trainer Application: In this demo we use a boxing punch detection example to show how simple it is to build application specific sports wearable algorithms with SensiML. The demo application as shown can provide accurate detection of jabs, uppercuts, hooks, and overhand punches making it very easy to generate a wearable boxing training application that can guide users on proper form and technique, count repetitions, timing, and duration of workout sessions. The results are collected by a miniaturized sensor module capturing 3-axis accelerometer and gyro data with results calculated locally in real-time on the sensor itself. Data sent wirelessly to a companion mobile application running on a tablet/phone includes just recognized punch type and feature extraction for the given model so user application integration is extremely simple.
In this manner, many similar ‘long tail’ sports wearable applications can be built quickly and efficiently using the SensiML Analytics Toolkit.
Demo #2 – Worker Ergonomic Box-Lifting Wearable: In this demo we show how SensiML is capable of detecting more subtle nuances in time-series motion data as well. In this case, we show how we can distinguish proper versus improper form for a user lifting a box. When you consider the number of warehouse, factory, and field workers who at risk for potential back injuries and worker’s compensation claims due to preventable poor lifting technique, the value of such an application becomes clear. With SensiML Analytics Toolkit, innovative wearable IoT device vendors can readily build accurate ergonomic models capable of distinguishing proper form for training and/or compliance. This application was built within two weeks using captured data and SensiML automated algorithm generation.