Part Number: AWRL6844
Hi TI Team,
Our company is currently developing an in-cabin System using the AWRL6844 platform.
We recently conducted data collection in a real vehicle across 5 different seating positions. For each position, we recorded:
One sample of an infant on the seat.
One sample of an infant in the footwell area.
One sample of an adult.
This resulted in a total of 15 data samples.
Attached are our training results. As you can see, the model achieved high accuracy during the offline training phase. However, after integrating the generated coefficients (classifier_weights.h) into the radar firmware, we observed a significant performance gap. In real-time testing, the radar seems only capable of distinguishing between an Empty Seat and Occupied (Life presence), failing to accurately classify "Infant" vs. "Adult" as the training results suggested.
Could you please provide insights into the following:
Another question，In the training featset, is the SNR value processed with a log transformation twice?
We look forward to your professional guidance.
Best regards,
Henry Lin
