Research on the Fusion Detection Method of Extreme Learning Machines and Sparse Coding

In the era of vast data availability, extracting crucial information and achieving accurate classification is a significant challenge, particularly in industrial settings where data is often noisy and redundant. To address this issue, the present study proposes a novel and more robust classifier by enhancing the performance of the extreme learning machine (ELM) and leveraging sparse coding techniques. The ELM is optimized to effectively handle complex data environments, and sparse coding techniques are incorporated to facilitate the extraction of essential features from raw data. By combining these features with the improved ELM, a classifier that is both more robust and efficient in its classification capabilities is constructed. Experimental validation on two datasets, the corona current dataset and the Chinese Power Line Insulator Dataset (CPLID), demonstrates that the proposed method exhibits exceptional anti-interference ability, high-speed, and accuracy when dealing with high-dimensional, noise-contaminated, redundant, and complex data. This study successfully develops a novel and efficient classification method by synergistically combining the improved ELM with sparse coding techniques. This method demonstrates exceptional performance in image processing and time-domain current signal classification tasks, and it has the advantage of being lightweight and is more suitable for rapid analysis of massive data.