Drowsiness behind-the-wheel is one of the major causes of road accidents leading to severe injuries and an increased number of deaths. A quick-responding and a high precision detection system, which can determine the state of drowsiness and alert the driver before it’s too late, is required to control the drowsiness levels while driving. The higher efficiency of the physiological signals over the behavioral signals aids the early detection of drowsiness as monitoring physiological signals is more effective than observing the behavior of the driver. In this thesis, a unique system, which uses an Electrocardiogram (ECG) to compute Heart Rate Variation (HRV) tuned with machine learning, is developed and tested to determine the state of drowsiness. Many researchers have observed noteworthy variations in the power spectra of Low Frequency and High Frequency regions when a person undergoes a transition from the state of being alert to that of being drowsy. An advanced algorithm is developed and tuned with machine learning once the feature extraction from HRV is accomplished with the help of the Wavelet Transform (WT). The whole system is implemented, and it is observed that accuracies as high as 90-95% are achieved with a detection time of less than a minute. It is concluded that a successful design of such a system assists in the early detection of drowsiness with high accuracies and will help reduce the number of casualties due to road accidents.