Genome sequences are highly complex in structure and their analysis and classification is a crucial challenge. Several machine learning and deep learning techniques were implemented for the classification of various sequences associated with various species. Classification of various regions of the human genome was attempted by only a few studies. In this study, the classification of the human genome was performed. In specific, three regions of the human genome namely coding, long non-coding, and pseudogenes were classified using deep learning models. Raw data was collected from NCBI, RNA central and pseudogene.org open-source website. Several pre-processing techniques were employed for the sequences to be fed to deep learning algorithms. One of the issues in handling genomic sequences is preserving the features of the sequence. For this purpose, Voss representation is used. In this study, two models CNN and hybrid CNN – Bidirectional GRU were employed. Seven models were generated using two models by tuning the parameters. The models are evaluated using accuracy and training time. From the experimental results, hybrid CNN – Bidirectional GRU has achieved the highest validation accuracy of 0.9431 with a better fitting curve.