The ability of Massive Multiple-Input Multiple-Output (Massive MIMO) technology to provide excellent spectral efficiency and increased network capacity makes it an essential enabler for the next generation of wireless communication systems. However, the intricacies of high-dimensional channel matrices, pilot contamination, and sparse multipath propagation settings make accurate channel estimation in Massive MIMO systems a major issue. Conventional estimating techniques, such as compressed sensing and Least Squares (LS), are more computationally demanding and have worse accuracy when the signal-to-noise ratio (SNR) is low. In order to address these issues, this research presents a hybrid framework for sparse channel estimation that makes use of deep learning, particularly Convolutional Neural Networks (CNN) and Recurrent Neural Networks (RNN). The suggested framework is tested under Rayleigh fading conditions in a Massive MIMO-OFDM setting. Metrics like Bit Error Rate (BER), Mean Square Error (MSE), and estimation accuracy are used in performance evaluations. When compared to conventional LS and compressed sensing techniques, simulation results show that the CNN-RNN methodology considerably lowers BER and improves channel estimation performance. Additionally, the suggested paradigm strengthens robustness in dynamic wireless environments and successfully reduces pilot overhead. For the advanced wireless communication systems of 5G and the upcoming 6G, the created framework provides an effective and scalable solution.