Assessing alcohol intoxication against physiological signs is an emerging field of investigation, and has potential uses both in health surveillance and safety. Therefore, this thesis explores the performance classification of the user from photoplethysmographic (PPG) signals obtained from wrist-worn sensors to detect sobriety versus alcohol intoxication. The main aim of this work is to investigate the feasibility of PPG and its frequency-domain features, including heart rate variability (HRV), for the detection of alcohol. A dataset of 55 subjects was collected, including PPG signals recorded under both sober and intoxicated states. Preprocessing techniques applied on the raw PPG data, including wavelet-based denoising, baseline correction and normalization were performed to improve the quality of the signal used in feature extraction.