Article: The analysis of algorithmic composition of polyphonic Miao folk songs using deep learning Journal: International Journal of Arts and Technology (IJART) 2026 Vol.16 No.9 pp.201 - 227 Abstract: polyphonic folk songs are characterised by independent vocal parts, interwoven melodies, and oral transmission. They constitute an important component of ethnic music culture and intangible cultural heritage. Due to a declining number of inheritors, insufficient digitisation of musical repertoires, and the difficulty of annotating polyphonic structures, traditional automatic composition methods struggle to simultaneously preserve Miao musical style, model polyphonic relationships, and extend application scenarios. This study focuses on the automatic generation of Miao polyphonic folk songs and proposes an algorithmic composition framework that integrates environmental perception data with deep learning models. The system converts environmental parameters into musical control variables, including tempo, mode, dynamics, and harmonic density. It adopts a hierarchical generative architecture composed of a bidirectional gated recurrent neural network (Bi-GRNN) and a transformer model. The findings indicate that the proposed environment-aware polyphonic generation approach preserves the essential stylistic characteristics of Miao folk songs. Inderscience Publishers - linking academia, business and industry through research

Title: The analysis of algorithmic composition of polyphonic Miao folk songs using deep learning

Authors: Bin Liu; Ru Chen

Addresses: School of Music and Dance, Nanchang Institute of Technology, Nanchang, Jiangxi, 330044, China ' School of Electronic Information Engineering, Nanchang Institute of Technology, Nanchang, 330000, China; Faculty of Education and Languages, Higher Education Learning Philosophy University, Kuala Lumpur, 50490, Malaysia

Abstract: polyphonic folk songs are characterised by independent vocal parts, interwoven melodies, and oral transmission. They constitute an important component of ethnic music culture and intangible cultural heritage. Due to a declining number of inheritors, insufficient digitisation of musical repertoires, and the difficulty of annotating polyphonic structures, traditional automatic composition methods struggle to simultaneously preserve Miao musical style, model polyphonic relationships, and extend application scenarios. This study focuses on the automatic generation of Miao polyphonic folk songs and proposes an algorithmic composition framework that integrates environmental perception data with deep learning models. The system converts environmental parameters into musical control variables, including tempo, mode, dynamics, and harmonic density. It adopts a hierarchical generative architecture composed of a bidirectional gated recurrent neural network (Bi-GRNN) and a transformer model. The findings indicate that the proposed environment-aware polyphonic generation approach preserves the essential stylistic characteristics of Miao folk songs.

Keywords: bidirectional gated recurrent neural network; transformer model; algorithmic composition; polyphonic Miao folk songs; cultural tourism.

DOI: 10.1504/IJART.2026.155630

International Journal of Arts and Technology, 2026 Vol.16 No.9, pp.201 - 227

Received: 17 Apr 2026
Accepted: 23 May 2026
Published online: 07 Aug 2026 *

Title: The analysis of algorithmic composition of polyphonic Miao folk songs using deep learning

Keep up-to-date