Meditation is often regarded as an alternative therapeutic approach for managing psychological disorders, helping to alleviate symptoms of depression and anxiety. Heart rate variability (HRV) is one of the most effective techniques to figure out how well the heart is working. The aim of the study is to explore the use of machine learning methods combined with HRV analysis to detect physiological changes due to meditation, with potential applications in therapeutic interventions for managing psychological disorders. The Kubios HRV software was employed to extract HRV from the RR intervals of the Chi meditation electrocardiogram (ECG) dataset, both before and during meditation. This study assesses the accuracy of various classifiers in analyzing heart rate variability signals to differentiate between pre-meditative and meditative states. We selected the top 15 features from the extracted set using the Recursive Feature Elimination (RFE). Given the small sample size of the Chi meditation dataset (8 subjects), we applied a data augmentation technique that averages feature sets to address the risk of overfitting. Four classifiers are applied with augmented and non-augmented data to assess the efficacy of the technique: random forest, logistic regression, support vector machine, and k-nearest neighbor. The experimental results show a significant improvement in classifier performance for distinguishing between before-meditation and during-meditation states when using augmented data compared to non-augmented data. The k-nearest neighbor classifier outperforms the other models, achieving an average accuracy, recall, and F1-score of 99.5%.