Our Research Team Developed a More Efficient Audio Embedding Model