Today, creation of smart algorithms is largely the realm of large device OEMs who can justify the significant investment in firmware development, digital signal processing, data science, application programming and domain expertise. Even then, these large OEM teams must limit algorithm scope to fixed-function, must-have features as the process is neither well automated nor scalable. As examples, consider the fixed 'wake word' required to trigger smart home hubs or the fixed 'step' or 'stair climb' motion detection algorithms of consumer fitness trackers that ignore or misrepresent all other valid modes of exercise activity.
For the rest, namely mid/low volume applications and smaller product teams, without practical means for generating smart sensor algorithms at all, they are left with either cloud processing or companion smartphone applications as the only practical options despite the shortcomings these dependencies inject in application performance, latency, and user experience.
SensiML's automated learning algorithm toolkit addresses these challenges head on solving the time consuming and expertise demanding aspects of algorithm design.
For large teams, this translates to easily extensible datasets and expanded use cases allowing teams to not only offer far more functionality in the same product development timeline, but also opens up new business models where smart devices can become the vehicle for ongoing application services powered by learning and updateable sensor algorithms.
For smaller teams, SensiML's toolkit offers the freedom to build true smart sensors without over-reliance on smartphone processing or cloud analytics that constrain your user requirements. Smaller teams are empowered to create complete algorithms with focus on their specific domain expertise and only average programming skills.