Use of autocorrelation analysis to characterize audibility of low-frequency tonal signals

Abstract

Autocorrelation function (ACF) parameters were used to identify low-frequency tonal sound detected in actual living environments. Five houses whose residents had made complaints for unidentified noise were selected as measurement sites. The sounds and the residents' detection responses were recorded simultaneously inside a room in each house. When they heard the suspected noise, the participants pushed a response button on a portable recording device as the sound was recorded. Results showed that tonal components in the low-frequency range were highly correlated with the sound detection. This study suggests that autocorrelation analysis can reveal the human detection of low-frequency tonal signals. Low-frequency tonal components were identified and quantified using ACF parameters: the delay time and amplitude of the ACF's first dominant peak. The amplitude was useful to describe the detection and prominence of low-frequency tonal components in noise.