The cognitive radio (CR) technology seeks to improve the harness of the radio spectrum with spectrum sensing as one of the important parts. But, as mentioned earlier, it is quite difficult to achieve accurate spectrum sensing because system performance is affected by various enabling factors. Artificial intelligence (AI) and deep learning (DL) allow new spectrum sensing techniques to be formulated based on new possibilities. In this paper, a model based on the connection between the models of recurrent neural network (RNN) and bidirectional long short-term memory (BILSTM) has been proposed for spectrum sensing. This approach uses the combined feature extraction function of both the RNN and BILSTM networks data models. The RNN is used for spatial feature extraction, and the BILSTM network is used for temporal features. Thus, unlike some other serial network connections, this parallel connection enables processing the original dataset directly and does not eliminate the information features. Experimental results depict that the proposed approach has better accuracy compared to the conventional cooperative detection algorithms attained the detection probability can reach more than 90% under low SNR conditions when 9 cooperative users and 10 transmit powers.