A multimodal dataset for electric guitar playing technique recognition.

Area of Science:

• Music Technology
• Machine Learning
• Signal Processing

Background:

• Automatic detection of musical instrument playing styles is crucial for intelligent music coaching software.
• Current methodologies are limited by the availability of comprehensive, real-world datasets for training machine learning models.
• Existing datasets often lack the diversity and completeness required for robust playing technique identification.

Purpose of the Study:

• To introduce a novel multimodal dataset of electric guitar playing techniques.
• To facilitate the development of advanced machine learning models for automatic guitar playing style assessment.
• To provide a valuable resource for creating intelligent music coaching and training applications.

Main Methods:

• A multimodal dataset comprising 549 video (MP4) and audio (WAV) samples of nine electric guitar techniques was created.
• Recordings were made using a smartphone, simulating real-world conditions with diverse exercises, guitars, and amplifier simulations.
• Accompanying musescores were provided, and Support Vector Machine (SVM) and Convolutional Neural Network (CNN) models were developed for technique classification using audio data.

Main Results:

• The dataset successfully supported the development and testing of SVM and CNN models for classifying electric guitar techniques.
• The multimodal nature and real-world recording setup enhance the dataset's utility for practical applications.
• Experimental results demonstrated the feasibility of using the dataset for training effective machine learning models.

Conclusions:

• The introduced multimodal dataset addresses the scarcity of resources for electric guitar playing style analysis.
• This dataset is a significant contribution to the field of intelligent music education and performance analysis.
• The publicly available code and extendable dataset format encourage further research and development in automated music technique recognition.