Abstract This paper presents an AI driven digital music creation system based on recursive neural networks (RNNs), designed to automate and enhance the music composition process. By leveraging RNNs particularly architectures like LSTM the system learns patterns from MIDI training data to generate new musical pieces that reflect specific styles, rhythms, and harmonic structures. The approach addresses the complexity and unpredictability... This paper presents an AI driven digital music creation system based on recursive neural networks (RNNs), designed to automate and enhance the music composition process. By leveraging RNNs particularly architectures like LSTM the system learns patterns from MIDI training data to generate new musical pieces that reflect specific styles, rhythms, and harmonic structures. The approach addresses the complexity and unpredictability of traditional digital music creation by reducing reliance on manual input and musical expertise. The study outlines the neural network’s architecture, training process, and the use of activation functions, such as Sigmoid, alongside feature selection and data preprocessing techniques to enhance performance. Experimental results show that processed data yields an 81.2% classification accuracy, rising to 90% with paragraph based feature methods. The system demonstrates strong potential in applications such as music education and live performance, though challenges remain in expressing emotional depth and ensuring creative diversity. Future work includes integrating large language models, such as ChatGPT, to enable expressive and improvisational AI musicians. Read More Read Less

Abstract This paper presents an innovative method for digitizing and analyzing sports training trajectory data using the mean shift algorithm combined with color histogram features for robust motion tracking. The authors model the human body as a 16 joint skeletal framework, representing joint rotations via quaternions mapped in to space to avoid singularities associated with Euler angles. To manage high... This paper presents an innovative method for digitizing and analyzing sports training trajectory data using the mean shift algorithm combined with color histogram features for robust motion tracking. The authors model the human body as a 16 joint skeletal framework, representing joint rotations via quaternions mapped in to space to avoid singularities associated with Euler angles. To manage high dimensional motion capture data, they apply a linear time invariant system for dimensionality reduction. The mean shift algorithm chosen for its non parametric nature and resilience to object deformation, rotation, and speed variations is used to track athletes’ motion paths by leveraging consistent color features across video frames. Experimental results demonstrate that this approach outperforms both Space Mocap and MATLAB based trajectory methods in accuracy and error rate. The study also integrates Dynamic Time Warping (DTW) to compare motion sequences, showing that intra class motion distances are significantly smaller than inter class ones, thereby enabling effective behavior classification. Motion sequences are categorized into dynamic (e.g., jumps, squats) and static (e.g., lifting, waving) actions. The system achieves precise segmentation and recognition of human activities, even under conditions of occlusion or changes in posture. Supported by motion capture databases and validated through comparative analysis, the proposed technique offers a reliable, parameterfree solution for real time athlete tracking and motion behavior analysis in sports science applications. Read More Read Less

Abstract This paper presents an electrical control system for a harvesting robotic arm that integrates Programmable Logic Controller (PLC) technology with the Particle Swarm Optimization (PSO) algorithm. Aimed at addressing labor shortages and enhancing harvesting efficiency in modern agriculture, the system uses PLC for reliable, real time control of the robotic arm’s movements. At the same time, PSO optimizes the harvesting... This paper presents an electrical control system for a harvesting robotic arm that integrates Programmable Logic Controller (PLC) technology with the Particle Swarm Optimization (PSO) algorithm. Aimed at addressing labor shortages and enhancing harvesting efficiency in modern agriculture, the system uses PLC for reliable, real time control of the robotic arm’s movements. At the same time, PSO optimizes the harvesting sequence to minimize travel distance and positioning errors. The control architecture includes sensor inputs (X1–X4) and electromagnetic valve outputs (Y1–Y4), with image processing for fruit recognition and localization. Experimental results demonstrate that the PSO enhanced system significantly improves harvesting accuracy, reduces errors especially under challenging conditions like intense light and increases overall operational efficiency. The approach enables precise, adaptive fruit picking while preserving produce quality, showcasing the potential of intelligent algorithms combined with industrial automation in agricultural robotics. This synergy offers a scalable solution for sustainable, high efficiency harvesting in diverse crop environments. Read More Read Less