Article: A music style recognition system by multi-task LSTM and its application in Orff music pedagogy Journal: International Journal of Reasoning-based Intelligent Systems (IJRIS) 2026 Vol.18 No.19 pp.51 - 62 Abstract: In order to solve the bottleneck problems of subjective evaluation and lagging teaching feedback in traditional music education, this study proposes a bidirectional long-short-term memory (BI-LSTM) multi-task learning framework that integrates harmonic/percussive source separation (HPSS) preprocessing technology. The system provides real-time quantitative teaching feedback according to students' performance by decoupling the characteristics of melody flow and rhythm flow. The international free music archive (FMA) dataset was used for technical benchmarking, and a three-week Orff teaching intervention experiment was conducted for 60 primary school participants. The experimental results showed that the system achieved 94.5% F1 score. Verified by analysis of variance (ANOVA), the accuracy of the students' discrimination of music styles was significantly improved from the initial benchmark value of 65.4% to 91.2% after intervention. The model system proposed in this study provides an objective and stable technical paradigm and decision support for the interactive mode of intelligent music education. Inderscience Publishers - linking academia, business and industry through research

Title: A music style recognition system by multi-task LSTM and its application in Orff music pedagogy

Authors: Shanshan Huang

Addresses: School of Music and Dance, Yulin Normal University, Yulin 537000, Guangxi, China

Abstract: In order to solve the bottleneck problems of subjective evaluation and lagging teaching feedback in traditional music education, this study proposes a bidirectional long-short-term memory (BI-LSTM) multi-task learning framework that integrates harmonic/percussive source separation (HPSS) preprocessing technology. The system provides real-time quantitative teaching feedback according to students' performance by decoupling the characteristics of melody flow and rhythm flow. The international free music archive (FMA) dataset was used for technical benchmarking, and a three-week Orff teaching intervention experiment was conducted for 60 primary school participants. The experimental results showed that the system achieved 94.5% F1 score. Verified by analysis of variance (ANOVA), the accuracy of the students' discrimination of music styles was significantly improved from the initial benchmark value of 65.4% to 91.2% after intervention. The model system proposed in this study provides an objective and stable technical paradigm and decision support for the interactive mode of intelligent music education.

Keywords: long short-term memory neural network; music style recognition system; Orff music pedagogy; deep learning technology; recurrent neural network.

DOI: 10.1504/IJRIS.2026.155635

International Journal of Reasoning-based Intelligent Systems, 2026 Vol.18 No.19, pp.51 - 62

Received: 09 Apr 2026
Accepted: 25 May 2026
Published online: 07 Aug 2026 *

Title: A music style recognition system by multi-task LSTM and its application in Orff music pedagogy

Keep up-to-date