Multidimensional acoustic feature fusion with attention-guided light weighted CNN for improved speaker recognition | Synapse