Language aims to have effective handwritten recognition which purpose is writer involvement on written documents concerned with the Non-Latin language in Bangla. In the Non-Latin language, Bangla claims the necessity of Auto-Learned Features with a sufficient manner supported by Human Constructed Features. According to this problem assertion of handwritten documents, security is sought for the proposal of document identification. In a sense, the Bangla language has a large vocabulary where it renders so many complex words including compound characters. On account of the Bangl.a language, usable boundaries are large because of the volume of manpower. Therefore, handwritten documents arise a great number a couple of times. Addressing those features of the problem proposed a model with CNN where it outperformed the handwritten documents in Bangla that mixed with different writers wrote the same documents in a different way. Moreover, the SVM comparison strongly stated that CNN is an effective approach on account of 70% training and 30% testing validation concern with 83%.