In Global Navigation Satellite Systems (GNSS), receivers are susceptible to various types of jammers. Detecting these jammers is crucial for developing effective antijamming techniques due to the increasing complexity of the communication environment and the proliferation of jamming technologies. Different kinds of jammers exhibit distinct behaviors in both the time and frequency domains, leading to various forms of degradation in GNSS receiver performance. This study investigates the detection of two specific kinds of jammers: continuous wave jammers and chirp jammers. Consequently, extracting the original signal or accurately determining the features of the jamming signal using conventional techniques can pose significant challenges. This article introduces a signal jamming detection deep learning model with a feature selection algorithm. After preprocessing the dataset and analyzing the GNSS data, a feature selection algorithm using the principal component analysis (PCA) combined with the Bayesian optimization (BO) algorithm is proposed, which processes the data sequentially in the following leads in a systematic manner and distinguishes between two unique signal categories: normal signals and jamming signals. Then, bidirectional long short-term memory with attention mechanism (BiLSTM-A) is applied to detect the jamming signals. Numerical results and a confusion matrix validate the correctness and efficiency of the proposed PCA-BO feature selection algorithm. The results of the BiLSTM-A model demonstrate 98.95% accuracy. Moreover, PCA-BO results in a $\mathbf {33}\%$ dimensionality reduction and the learning time is reduced by $\mathbf {23}\%$, with nearly the same accuracy.