Audio-Visual Active Speaker Identification: A comparison of dense image-based features and sparse facial landmark-based features