COVID-19 has significantly impacted global healthcare, emphasizing the need for rapid and non-invasive diagnostic tools. This study proposes a machine learning (ML)-based approach for classifying COVID-19 cases using vowel /a/ sound recordings. The statistical features are extracted using the bilinear transform-based Warp Filter Bank Cepstral Coefficients (BWFCC), along with their first and second-order derivatives, using vowel sound signals. In the BWFCC, the warp parameters are adjusted to generate different filter variations, enabling the identification of the optimal warp parameter value that best distinguishes COVID-19-positive cases from healthy individuals. These features are then fed into ML algorithms, including Logistic Regression (LR), Support Vector Machine (SVM), and Multi-Layer Perceptron (MLP) for classification. The experimental results demonstrate that the proposed methodology effectively discriminates between COVID-19-positive and healthy individuals, highlighting the importance of optimizing the warp filter bank parameters ( \( \alpha \) ) in improving accuracy. For the 0.75 \( \alpha \) value, MLP achieved the highest accuracy of 83% using oversampling and 81% without sampling. This study shows vowel /a/ sound effectively capture COVID-19 biomarker to distinguish the signal. This approach offers a promising, noninvasive, cost-effective, and scalable diagnostic tool for COVID-19 detection using vowel sound analysis.