At present, artificial intelligence technology represented by deep learning has been applied to various industries, and it also plays a huge influence in the field of magnetic resonance. Deep learning method has been proved to be effective in spectral signal reconstruction, denoising and recognition. In terms of signal reconstruction, sparse sampling can be used to reduce the time of data acquisition, and the missing sampling points can be reconstructed by deep learning method, so as to restore the complete spectrum. In the aspect of signal denoising, the deep learning method can remove the noise in the magnetic resonance signal and improve the signal-to-noise ratio. In terms of signal recognition, the deep learning model can determine the position and intensity of spectral peak by learning the pattern and feature of peak....... In short, deep learning promotes the efficiency of magnetic resonance spectrogram processing, and has important significance in extracting spectral information effectively and improving the quality of the spectrogram.