听觉脑电图解码是大脑解码中的一个重要的分支，可以从声音刺激下的神经活动中获取和解码相应信息，从而重建出大脑对听觉信息近似的反映。立体脑电图（SEEG）是通过植入大脑皮层的深入电极测量大脑皮质和皮质下活动的方法。相关研究表明，音乐能够提高大脑特定区域的活动水平，从而对人的情绪产生影响，因此在音乐刺激下从大脑活动中解码其诱发的情绪对于大脑信号的解码有着理论和实践意义。本文在不同音乐刺激下监测到的SEEG数据的基础上，使用多种深度学习方法将SEEG数据与对应的音频情绪表征构建映射关系，解码音乐刺激诱导的大脑情绪活动类型并比较模型基础性能。同时使用不同数据变换及映射方法处理数据，提升模型的学习和理解能力，研究将音乐刺激下立体脑电图与其诱发的情绪联系起来的最佳模型。

Auditory EEG decoding is an important branch of brain decoding, which can obtain and decode the corresponding information from the neural activities stimulated by sound, so as to reconstruct the brain's approximate response to auditory information. Stereo electroencephalography (SEEG) is a method of measuring activity in the cerebral cortex and subcortical through deep electrodes implanted in the cerebral cortex. Relevant studies have shown that music can increase the activity level of specific areas of the brain, thereby affecting people's emotions, so decoding the emotions induced by brain activities under music stimulation has theoretical and practical significance for the decoding of brain signals. Based on the SEEG data monitored under different music stimuli, this paper uses a variety of deep learning methods to construct a mapping relationship between the SEEG data and the corresponding audio emotion representations, decode the types of brain emotional activities induced by music stimuli, and compare the basic performance of the model. At the same time, different data transformation and mapping methods were used to process the data to improve the learning and comprehension ability of the model, and the best model to associate stereoscopic EEG with the emotions induced by music stimulation was studied.