Article: Multi-style music generation and sound quality enhancement technology based on MC-DCGAN model Journal: International Journal of Information and Communication Technology (IJICT) 2026 Vol.27 No.41 pp.70 - 91 Abstract: With the rapid development of artificial intelligence, generative adversarial networks have been widely applied to music generation. However, existing methods still face limitations in multi-style control, temporal coherence, and sound quality enhancement. To address these issues, this study proposes a multi-style music generation and sound quality enhancement approach based on an improved deep convolutional generative adversarial network. The method integrates a multi-condition control mechanism, a temporal structure generator, and an adaptive instance normalisation module to improve melody coherence, style controllability, and audio quality. Experimental results show that the proposed method achieves a style classification accuracy of 89.4% and a cross-section coherence of 0.83, with precision and recall of 0.74 and 0.69, respectively. For popular music styles, the generation accuracy approaches 98%, note diversity exceeds 95%, rhythm consistency reaches 0.98, and cross-bar coherence reaches 0.97. These results demonstrate the effectiveness and robustness of the proposed method. Inderscience Publishers - linking academia, business and industry through research

Title: Multi-style music generation and sound quality enhancement technology based on MC-DCGAN model

Authors: Junyan Wang

Addresses: Faculty of Music, Silpakorn University, Bangkok, 10170, Thailand

Abstract: With the rapid development of artificial intelligence, generative adversarial networks have been widely applied to music generation. However, existing methods still face limitations in multi-style control, temporal coherence, and sound quality enhancement. To address these issues, this study proposes a multi-style music generation and sound quality enhancement approach based on an improved deep convolutional generative adversarial network. The method integrates a multi-condition control mechanism, a temporal structure generator, and an adaptive instance normalisation module to improve melody coherence, style controllability, and audio quality. Experimental results show that the proposed method achieves a style classification accuracy of 89.4% and a cross-section coherence of 0.83, with precision and recall of 0.74 and 0.69, respectively. For popular music styles, the generation accuracy approaches 98%, note diversity exceeds 95%, rhythm consistency reaches 0.98, and cross-bar coherence reaches 0.97. These results demonstrate the effectiveness and robustness of the proposed method.

Keywords: multi-style music generation; generate adversarial networks; deep convolutional networks; sound quality enhancement; temporal structure generator; TSG; adaptive instance normalisation.

DOI: 10.1504/IJICT.2026.153375

International Journal of Information and Communication Technology, 2026 Vol.27 No.41, pp.70 - 91

Received: 13 Oct 2025
Accepted: 05 Dec 2025
Published online: 06 May 2026 *

Title: Multi-style music generation and sound quality enhancement technology based on MC-DCGAN model

Keep up-to-date