2021-04-29 06:30 PM
Hi~
I want to know the main difference between the 3 sensor parts.
I found in the datasheet that only the LSM6DSO32X is marked with machine learning cores.
Could this make any difference to the development or final product?
I finally include this sensor on a watch-type product and try to distinguish about 4 behavior patterns through machine learning.
Considering this, are there any more suitable parts?
Thank you~.
Solved! Go to Solution.
2021-04-30 01:34 AM
Hi @SKim.25 ,
>> I found in the datasheet that only the LSM6DSO32X is marked with machine learning cores.
Could this make any difference to the development or final product?
Yes, you are right. Part numbers with the final "X"- are provided with embedded Machine Learning Core.
Whether it could make difference, it depend on your application: if you are using the sensors just to acquire data that are then sent to the application processor for the machine learning (for example using neural networks), it is enough to use an IMU without MLC, so LSM6DSOP and LSM6DSO32 are suitable. You should choose the LSM6DSO32 in the case you need to capture signals up to 32g.
On the other side, if you want to optimize the consumption of your application, you should use embedded MLC, and so the LSM6DSO32X: this solution can be particularly suggested in the case the behavior pattern you have to detect are not so complex, i.e. can be defined by a small number of features and can be classified using decision tree models.
You can find some C code examples explaining the use of the MLC (on LSM6DSOX) on Github: the two cases that might be of more interest for you should be the Activity recognition for wrist and the Head gestures.
If my reply answered your question, please click on Select as Best at the bottom of this post. This will help other users with the same issue to find the answer faster.
-Eleon
2021-04-30 01:34 AM
Hi @SKim.25 ,
>> I found in the datasheet that only the LSM6DSO32X is marked with machine learning cores.
Could this make any difference to the development or final product?
Yes, you are right. Part numbers with the final "X"- are provided with embedded Machine Learning Core.
Whether it could make difference, it depend on your application: if you are using the sensors just to acquire data that are then sent to the application processor for the machine learning (for example using neural networks), it is enough to use an IMU without MLC, so LSM6DSOP and LSM6DSO32 are suitable. You should choose the LSM6DSO32 in the case you need to capture signals up to 32g.
On the other side, if you want to optimize the consumption of your application, you should use embedded MLC, and so the LSM6DSO32X: this solution can be particularly suggested in the case the behavior pattern you have to detect are not so complex, i.e. can be defined by a small number of features and can be classified using decision tree models.
You can find some C code examples explaining the use of the MLC (on LSM6DSOX) on Github: the two cases that might be of more interest for you should be the Activity recognition for wrist and the Head gestures.
If my reply answered your question, please click on Select as Best at the bottom of this post. This will help other users with the same issue to find the answer faster.
-Eleon