Temporal dynamics of pitch strength in regular interval noises

Area of Science:

• Psychoacoustics
• Auditory Perception
• Signal Processing

Background:

• Pitch strength of rippled noise (RN) and iterated rippled noise (IRN) is often modeled using the autocorrelation function's first peak height.
• Previous research established an exponential relationship between pitch strength and autocorrelation peak height for RN and IRN.

Purpose of the Study:

• To compare the pitch strengths and autocorrelation functions of rippled noise (RN) and a novel regular-interval noise, AABB.
• To investigate the discriminability and pitch strength differences between RN and AABB across various frequencies.
• To propose and validate a computational model explaining observed pitch strength variations.

Main Methods:

• Generated RN and AABB stimuli with identical pitch characteristics.
• Conducted listening experiments to assess the discriminability and perceived pitch strength of RN and AABB.
• Developed and tested a three-stage running-autocorrelation model incorporating a nonlinearity and temporal integration.

Main Results:

• AABB and RN were discriminable below approximately 250 Hz.
• AABB exhibited greater pitch strength than RN at pitches below 250 Hz; pitch strength was comparable above 250 Hz.
• When RN was replaced by IRN matched to AABB's pitch strength, discriminability was lost.

Conclusions:

• A significant pitch-strength difference exists between AABB and RN at low frequencies, explained by a proposed auditory model.
• The three-stage running-autocorrelation model, including temporal integration and a nonlinearity, accurately predicts pitch strength matching data.
• Auditory pitch perception involves complex temporal integration mechanisms beyond simple autocorrelation peak height.