Table 2.

Feature extraction hyperparameters for both accelerometer and audio signals. For accelerometer, 16, 32, 64 samples i.e. 160, 320 and 640 msec long frames overlap in such a way that the number of these frames i.e. segments (5 and 10) are the same for all events in our dataset. Similarly for audio signals, MFCCs are varied between 13 and 65 & frames are varied between 256 samples (11.61 msec) and 4096 samples (185.76 msec) in such a way that the number of these extracted frames are varied between 50 to 150, fixed for all events in out dataset.

FEATURE EXTRACTION HYPERPARAMETERS
Accelerometer Hyperparameters		Values
Frame ($\mathrm{\Psi }$)	Frame-length in samples, used to extract features	$2^k$ where $k=4,5,6$
Segments (C)	Number of frames extracted from the event	5, 10
Audio Hyperparameters		Values
MFCC ($\mathcal{M}$)	Number of lower-order MFCCs to keep	$13\times k$, where $k=1,\cdots ,5$
Frame ($\mathcal{F}$)	Frame-length in samples, used to extract features	$2^k$ where $k=8,\cdots ,12$
Segments ($\mathcal{S}$)	Number of frames extracted from the event	$10\times k$, where $k=5,7,10,12,15$