The greater rate of transmission is the most advanced features of significantly utilized in wireless communication system such as Multiple Input Multiple Output with Orthogonal Frequency Division Multiplexing (MIMO-OFDM). Nevertheless, an effective Channel State Information (CSI) is complex to acquire because of the large number of feedback overhead impacted through numerous antennas. Hence, this research proposes the Deep Learning (DL) approach Backpropagation based Deep Neural Network (BP-DNN) approach for channel estimation in MIMO-OFDM. An introduced approach performs the feedforward framework that effectively comprehends an estimation, redevelopment and the compression of the downlink channels in MIMO. The various networks are designed to work on the estimation and the feedback frameworks like implicit and explicit. The explicit network acquires the Signal-to-Noise-Ratio (SNR) to obtain effective outcome, while implicit straightly compresses the pilot and received to minimize network parameters. The experimental results shows that the proposed BP-DNN approach effectively attains the Mean Square Error of 0.0019 at the 0th dB SNR as compared to the existing method like CNN Convolutional Neural Network based Auto Encoder (CNN-DE).