<p>Precise assessment of hearing threshold is critical in clinical and forensic settings. 40 Hz Auditory Evoked Response Potential (AERP) threshold evaluation is an objective electrophysiological technique for estimation of hearing threshold, particularly in non-cooperative individuals. However, it still relies on physician interpretation of AERP waveforms, leading to variability and limiting large-scale, automated assessments. To address these problems, we propose the Self-attention Enhanced dual-stream Network with Gramian Angular field (SENGA) for classifying 40 Hz AERP waveforms and determining its response thresholds. First, Gramian angular field (GAF) transformation is employed to convert one-dimensional AERP waveform signals into two-dimensional images, preserving temporal information. Then, a dual-stream self-attention enhanced network is proposed to simultaneously extract complementary features from the GAF transformed images including the Gramian angular summation field (GASF) and Gramian angular difference field (GADF) images, as the two input of the dual streams to integrate both global and local information. The self-attention mechanism is utilized in the network to enhance the model’s ability to focus on key feature areas and thus improve assessment accuracy. Experimental results show that the proposed method achieves an accuracy of 0.9450 in the 40 Hz AERP waveform classification and an accuracy of 0.8663 in the 40 Hz AERP response threshold determination. Our proposed method provides an objective and accurate tool for hearing threshold assessment in clinical and forensic medicine.</p>