Competitiveness in consumer sports and fitness wearables has led to a wealth of devices seeking to appeal to user interests with greater insight, more relevance to specific activities, and broader usefulness. The ubiquitous digital pedometer has given way to wearable devices for weightlifting, cycling, running, swimming and a variety of other activities. Insights are extending from basic “quantified self” metrics to form and technique monitoring.
High degree of user-to-user variation means either poor models or sizable algorithm development efforts to address corner cases
Algorithm development by hand-coding is not scalable yield niche applications
Devices quickly outliving their ability to drive new user insight leads to abandonment
The current model for building wearable insight algorithms is badly flawed and has necessarily led to a fragmented ecosystem of purpose-built devices. Just as PCs are general purpose platforms, wearable devices ideally should have similar extensibility to adapt and refresh with user needs, personalization, and application. SensiML addresses one of the key barriers in this process, the scalable development of application specific algorithms based on a common set of physical sensors. By removing the data science and coding requirements, SensiML focuses the algorithm development effort on collecting well labeled datasets across user types, activities, and an ever-expanding set of useful insights.
Scalable algorithm development workflow
Focus is on data collection and application domain expertise, not data science and firmware optimization
Dataset portability across hardware and sensor designs preserves effort and IP built into curated application datasets
With years of expertise in wearable algorithm design, SensiML has sample datasets available to demonstrate how its data-centric endpoint AI toolkit can drive better, faster, broader algorithms for wearable devices.