Early detection of neurocognitive and mental health alterations remains limited by the high costs and invasive nature of many traditional clinical protocols. In this context, vocal biomarkers have emerged as a non-invasive, low-cost, and accessible alternative capable of reflecting underlying cognitive and emotional states. Humming, as a source of vocal biomarkers, offers advantages over speech-based methods, including language independence, scalability, and privacy preservation. With the objective of evaluating whether vocal biomarkers extracted from humming are sensitive to aging-related cognitive changes, participants aged between 18 and 80 years completed a humming task under different levels of cognitive load (immediate humming vs. delayed reproduction). Participants repeated two different melodies; for each melody, participants performed six trials, three sung immediately and three sung after playback, resulting in twelve trials per participant. A range of prosodic and spectral voice features, commonly used in speech-based biomarker research, were extracted and analyzed, including measures related to pitch variability, temporal stability, and spectral energy distribution. Results revealed systematic variations in these spectral and prosodic features associated with age and cognitive performance, suggesting that humming captures meaningful information related to cognitive functioning across the lifespan. This highlights the potential of humming-based vocal biomarkers as a promising tool for monitoring age-related cognitive changes and for the early detection of neurocognitive alterations in both research and clinical settings.