Feature Fusion Based Audio-Visual Speaker Identification Using Hidden Markov Model under Different Lighting Variations

Table 1

Performance measurements among various combinations of appearance based, shape based, and appearance-shape based visual features and MFCC based, LPCC based, and MFCC-LPCC based audio features according to various SNRs.

SNR (dB)	Appearance based feature (%)	Shape based feature (%)	Appearance-shape based feature fusion (%)	MFCC based feature (%)	LPCC based feature (%)	MFCC-LPCC based feature fusion (%)
0	85.00	88.00	93.00	5.00	4.00	8.33
5	85.00	88.00	93.00	10.67	8.33	13.33
10	85.00	88.00	93.00	21.33	18.67	28.33
15	85.00	88.00	93.00	35.33	30.00	50.00
20	85.00	88.00	93.00	50.67	43.33	72.33
25	85.00	88.00	93.00	70.00	63.67	80.00
30	85.00	88.00	93.00	90.00	86.33	93.67
Average (%); (0 dB~30 dB)	85.00	88.00	93.00	40.42	36.33	49.42