<p>Quantum audio processing is of great significance for artificial intelligence-driven audio recognition technology. To address traditional deep learning methods’ low efficiency and insufficient temporal modeling capabilities in audio data processing, this paper proposes a quantum time-series encoding (QTSE) method and a quantum audio neural network (QANN). The QTSE method leverages entangled qubit sequences to efficiently encode audio data into quantum states while preserving essential features. Building on QTSE, a quantum audio neural network is further developed and applied to audio recognition and classification tasks through quantum circuit parameter optimization. Experimental results on the GTZAN dataset show that the proposed method achieves a classification accuracy of 75% under limited qubit conditions, outperforming traditional amplitude and angle encoding schemes. Furthermore, the proposed QANN architecture is inherently parameter-efficient, utilizing only 33 trainable parameters, a quantity substantially smaller than classical deep learning counterparts typically used for similar audio classification tasks. This research provides a new pathway for building efficient quantum audio processing technology and demonstrates broad application potential in quantum machine learning.</p>