Robust Sounds of Activities of Daily Living Classification in Two-Channel Audio-Based Telemonitoring

Table 4

Classification accuracy of promising system configurations, for all three microphone configurations of the noisy SDL database, across four different noise configurations.

Method	Radio normal	Rain	Radio loud	Talking	Average
	(Mic 1 and Mic 2)	(Mic 1 and Mic 2)	(Mic 1 and Mic 2)	(Mic 1 and Mic 2)
	(Mic 1 and Mic 3)	(Mic 1 and Mic 3)	(Mic 1 and Mic 3)	(Mic 1 and Mic 3)
	(Mic 1 and Mic 4)	(Mic 1 and Mic 4)	(Mic 1 and Mic 4)	(Mic 1 and Mic 4)
Baseline (channel with best SNR)	44.04%	61.39%	28.58%	39.54%	43.39%
	51.25%	61.71%	28.75%	44.61%	46.58%
	49.38%	64.91%	33.17%	44.42%	47.97%
SAD alone: CSLE	71.95%	97.18%	40.25%	69.46%	69.71%
	84.08%	99.77%	55.21%	79.67%	79.68%
	79.21%	98.66%	50.63%	73.13%	75.41%
ICASDL_Apriori + CSLE	71.63%	98.01%	41.04%	66.96%	69.41%
	83.46%	100%	55.88%	81.67%	80.25%
	77.88%	98.24%	48.71%	44.79%	67.41%
ICASDL_BestSNR + CSLE	73.04%	97.50%	41.08%	67.75%	69.84%
	82.79%	99.07%	57.17%	76.94%	78.99%
	76.63%	98.24%	49.00%	69.96%	73.46%
ICASSA + CSLE	71.04%	97.04%	41.58%	68.21%	69.47%
	82.50%	99.07%	55.29%	79.00%	78.97%
	75.00%	97.13%	47.29%	67.46%	71.72%