npj Biosensing

Table 2 Performance metrics of four classifiers (SVM, LDA, LR, and TREE) evaluated through leave-one-out cross-validation on the LGI-PPGI dataset.

From: Optimal signal quality index for remote photoplethysmogram sensing

Excellent vs. Unfit
	Index	N_SQI	E_SQI	K_SQI	P_SQI	Z_SQI	R_SQI	S_SQI	M_SQI
Overall	OF₁	0.77	0.75	0.75	0.51	0.5	0.49	0.45	0.27
SVM	SE	0.81	0.76	0.81	0.37	0.52	NaN	0.35	NaN
	PP	0.82	0.78	0.77	0.55	0.63	NaN	0.56	NaN
	F₁	0.81	0.76	0.79	0.54	0.7	NaN	0.53	NaN
LDA	SE	0.81	0.81	0.81	0.41	0.4	NaN	0.4	NaN
	PP	0.82	0.79	0.73	0.41	0.65	NaN	0.61	NaN
	F₁	0.81	0.79	0.77	0.62	0.61	NaN	0.61	NaN
LOGI	SE	0.76	0.76	0.74	0.37	0.3	NaN	0.35	NaN
	PP	0.81	0.85	0.79	0.56	0.62	NaN	0.61	NaN
	F₁	0.78	0.8	0.76	0.65	0.51	NaN	0.56	NaN
TREE	SE	0.69	0.64	0.74	0.5	0.45	0.5	0.46	0.29
	PP	0.67	0.67	0.61	0.55	0.41	0.49	0.47	0.26
	F₁	0.68	0.65	0.67	0.51	0.43	0.49	0.47	0.27

Excellent vs. Acceptable
	Index	N_SQI	E_SQI	K_SQI	S_SQI	M_SQI	Z_SQI	P_SQI	R_SQI
Overall	OF₁	0.72	0.69	0.68	0.67	0.67	0.66	0.64	0.59
SVM	SE	0.92	0.92	0.72	0.82	0.82	0.82	0.63	0.73
	PP	0.63	0.63	0.58	0.61	0.61	0.61	0.55	0.58
	F₁	0.75	0.75	0.63	0.7	0.7	0.7	0.59	0.65
LDA	SE	0.92	0.82	0.82	0.82	0.82	0.82	0.72	0.65
	PP	0.63	0.65	0.61	0.61	0.69	0.61	0.58	0.55
	F₁	0.75	0.72	0.7	0.7	0.75	0.7	0.63	0.59
LOGI	SE	1.00	0.82	0.82	0.82	0.82	0.82	0.82	0.65
	PP	0.66	0.65	0.61	0.61	0.69	0.61	0.61	0.55
	F₁	0.79	0.72	0.7	0.7	0.75	0.7	0.7	0.59
TREE	SE	0.55	0.57	0.73	0.63	0.43	0.55	0.75	0.45
	PP	0.6	0.57	0.67	0.55	0.5	0.51	0.57	0.63
	F₁	0.57	0.57	0.7	0.59	0.48	0.53	0.65	0.52

Acceptable vs. Unfit
	Index	K_SQI	P_SQI	N_SQI	Z_SQI	E_SQI	M_SQI	S_SQI	R_SQI
Overall	OF₁	0.68	0.67	0.57	0.5	0.4	0.36	NaN	NaN
SVM	SE	0.38	0	0.38	0.38	0	0	0	0
	PP	0.75	NaN	0.6	NaN	NaN	NaN	NaN	NaN
	F₁	0.75	NaN	0.67	NaN	NaN	NaN	NaN	NaN
LDA	SE	0.38	0	0.38	0	0.25	0	0	0
	PP	0.38	NaN	0.3	0	0.67	0.0	NaN	NaN
	F₁	0.75	NaN	0.67	NaN	0.57	NaN	NaN	NaN
LOGI	SE	0.25	0	0.13	0	0.13	0	0	NaN
	PP	0.67	NaN	0.17	NaN	0.5	0	NaN	NaN
	F₁	0.57	NaN	0.29	NaN	0.33	NaN	NaN	NaN
TREE	SE	0.38	0.25	0.38	0.25	0.13	0.25	0	0
	PP	0.3	0.5	0.3	0.25	0.17	0.14	0	0
	F₁	0.67	0.67	0.67	0.5	0.29	0.36	NaN	NaN

Binary classification was executed on rPPG signals derived using CHROM, GREEN, and OMIT methods within three distinct signal quality classes: Excellent, Acceptable, and Unfit. Reported metrics are Sensitivity (SE), Positive Predictivity (PP), and F₁ score (F₁). The N_SQI index was found to be most effective in Excellent vs. Unfit and Excellent vs. Acceptable comparisons, whereas K_SQI was predominant in the Acceptable vs. Unfit comparison. Note: OF₁ denotes the overall F₁ score, the mean of the F₁ scores across the four classifiers.
Bolded numbers indicate the highest classification rate achieved.

Back to article page

Search

Advanced search

Quick links