Which is the difference between FSM and MLC? And how to integrate the two features?

[Evelina CAMILLERI]

[webinar - Program decision trees in sensors with a Machine Learning Core]

[Evelina CAMILLERI]

FSM and MLC are two different features in our devices, although part of memory is shared between these two features. 

In general, the FSM is more suitable for gesture recognition based algorithms, as it can be programmed to recognize a certain pattern evaluating conditions sample by sample. Some examples of FSM applications are: wake-up and free-fall detection, flip up/down gesture recognition, tap/double-tap detection, position recognitions (e.g. face up/down, 4D, 6D, etc...).

On the other hand, MLC is more suitable for activity recognition based algorithms, or more in general when data patterns are periodic in time, since the MLC works with some features computed in a defined time window. Some examples of MLC applications are: human activity recognition (e.g. stationary, walking, jogging, driving, etc...), fitness activities recognition, motion intensity recognition or vibration monitoring, and also some position recognition.

FSM and MLC can be used independently, but can also work in parallel or together. When both FSM and MLC are used, a merge tool has to be used to generate a configuration including both features (MLC+FSM), since part of the device configuration is shared. This is possible with the "Merge" tool available in the Options tab of Unico GUI.

More details are included in the MLC application notes available on www.st.com/mems.