Combining Weighted Multi-resolution STFT Loss and Distance Fusion to Optimize Speech Enhancement Generative Adversarial Networks