Multimodal Authentication Token Through Automatic Part of Speech (POS) Tagged Word Embedding

The part of speech (POS) tagging token matching for individual’s identification is trending and transforming into one of the secured ways under multifactor authentication process wherein important features either of trait have to be necessarily match with the fused features vector of multiple traits. In this model, POS tagging used for token matching with the features of handwritten signature traits considering the technical feasibility of hidden Markov model (HMM) that’s a dual stochastic process suitable for simulation and processing of feature extraction, summation, concatenating functionalities, and performance comparison with RNN Bi-LSTM PoS tagged word embed base classifier on ResNet-50 architecture model for multifactor authentication to validate the multimodal model trending and transforming into reliable and secured way forward to develop robust and trustworthy system in near future. For example, IVRS call verification can be value added with word token POS tagging concatenating with features of handwritten signature and further adding more traits like iris, facial and finger print, i.e. image pattern base clustering with voice feature sampling. Such multimodal model will able to strengthening and preventing from cyber mischievousness and surely improve the off-line signature verification of the bank customer. The proposed automatic voice recognition model has further scope of extension for experimentation and developing the clusters between the samples of image and embodied voice features which may be emerged reliable validation approach for a multimodal robust system.