Can Arduino Nano BLE 33 sense run 2 TinyML?

Hardware Nano Family Nano 33 BLE Sense
1

I have two muscle sensors, one for the flexor and one for the bicep (arm). The EMG signals (Muscle Sensor) are feature extracted. There are many feature extraction methods, but let's consider the simple case where 1 EMG signal corresponds to 1 feature extraction. So, for the two EMGs (flexor and bicep), we have two feature extractions each.

I am using the Arduino Nano BLE 33 Sense. The Arduino processes the feature extraction from the flexor EMG as input, and the output is either a hand grasp or an open hand, using a classification model (assume SVM).

The feature extraction from the bicep EMG is combined with a load cell sensor and a Gyroscope MPU sensor as input for machine learning. The machine learning model here is assumed to be linear regression, with the output being the elbow angle ranging from 0 to 120 degrees.

The results from both ML models (regression and classification) then control a servo motor as an actuator. I'm using TinyML and my question is, can the Arduino Nano BLE 33 Sense run two ML models (regression and classification)? And if it can, is it capable of doing so in real-time?

2

The elbow angle is not related to the EMG signal.

3

from the biceps EMG signal + load cell sensor + gyroscope sensor into machine learning input to find the bending angle of the hand

4

Yes, but it does not give you an angle. You need 2x IMU to calculate the elbow angle. Gyro == first derivitive of angle + noise, and when you integrate it gives angle + noise*t + c.

You can prove that emg reading is not the angle: attach a prostetic lower arm to a test persons elbow so that it has an angle of ~ 10° - 20° to the underlying rasl arm. cover the "real" arm with tissue so that it is hidden. let the person sit and move the arm for a while (10 mins are enough). After that time the persions brain replaced mental image of the real arm with the fake arm - you can even cut it and the person will anticipate pain.

5

okay, later I will try 2 IMU, but back to the original question, can 2 TinyML machine learnings run together on an Arduino nano ble 33 sense?