Enhanced Sliding Discrete Fourier Transform (eSDFT) With Error-Bound Control for Real-Time Parallel Processing

Abstract

This study presents the enhanced Sliding Discrete Fourier Transform (DFT), a novel method for real-time frequency analysis optimized for multi-core platforms. Traditional SDFT approaches suffer from error accumulation and inefficiencies in parallel processing. The proposed eSDFT introduces a boundederror recursive formulation with m-sample shift updates, enabling efficient, scalable computation across CPU cores. A parallel update strategy ensures low-latency processing without compromising spectral accuracy. Complexity analysis reveals a significant reduction in operations per sample, while experiments on synthetic and real signals demonstrate that eSDFT achieves a speedup of up to 4 times compared to FFTW for N=8192, with negligible degradation in accuracy. The method is well-suited for high-throughput applications such as biomedical signal processing and vibration monitoring, offering both robustness and performance in constrained environments.