Perceptual–Acoustic Comparisons of Natural Sonic Environments: Applications for Nature-Deprived Populations

Abstract

As humanity becomes increasingly urbanized and disconnected from the nonhuman world, nature-based sonic environments are receiving increased attention from physicians and psychologists for their potential therapeutic attributes. Such benefits include helping to increase focus, speeding recovery from stress, and reducing healing time from trauma. Comparing, identifying, and understanding how and which sets of features from sonic environments hold therapeutic relevance however remains unclear. Mel (short for melody) frequency cepstral coefficients (MFCCs)—a popular digital signal processing (DSP) representation that engages auditory perception—may be helpful in addressing this need. In this article, we conduct a preliminary comparative analysis of four recorded natural environments by extracting perceptually salient sets of derivative feature signals from their audio tracks. These feature signals are widely used in the music and audio information retrieval community and move beyond the sonic dimensions illustrated in standard log-frequency spectrograms. We then apply principal component analysis to visually represent the relative importance of these feature sets in each environment's audio, demonstrating that perceptually linked MFCC features can play a significant role in signal-based discrimination between natural environments. To conclude, we discuss applications of DSP advances to research in acoustic ecology and nature-based sonic therapy and provide suggestions for future research that can inform short- and long-term interventions aimed at promoting psychological and physical healing.