ECG stands for Electrocardiogram and has played a pivotal role in early diagnosis of various cardiac ailments and thus has helped save lives and improve patient outcomes. Arrhythmia is irregular heartbeat. The proposed approach aims to introduce an alternative way of detecting arrhythmia, by using DenseKAN, an implementation of Kolmogorov-Arnold Networks (KANs) in tensorflow. KANs have limited parameters and computational requirements, making it lightweight compared to deeper architectures. Kolmogorov-Arnold Theorem serves as foundational principle for KANs. For the dataset, we have acquired database through Kaggle that contains 34 features of Lead II, V5 derived from MIT-BIH Arrhythmia Database and MIT-BIH Supraventricular Database totaling 285117 records and aims to classify signal into 5 classes. The proposed study compares 2 architectures – Single Layer DenseKAN and a hybrid architecture – a LSTM and DenseKAN. The study has resulted in 97% accuracy in hybrid architecture and 93% in single-layer DenseKAN and successfully classified 4 out of 5 classes.