Functional clustering of mouse ultrasonic vocalization data.

Area of Science:

• Animal Behavior
• Bioacoustics
• Computational Biology

Background:

• Mouse ultrasonic vocalizations (USVs) are crucial in scientific research.
• Analyzing USVs is challenging due to noise and diverse call patterns.
• Existing methods lack comprehensive automatic analysis capabilities.

Purpose of the Study:

• To develop and improve methods for noise reduction, USV call detection, and automatic clustering.
• To enable more complete and accurate analysis of mouse ultrasonic vocalizations.
• To provide a robust framework for studying USV communication in mice.

Main Methods:

• Implemented advanced noise reduction techniques for raw USV signals.
• Utilized B-spline basis functions to characterize USV calls as functional data.
• Employed a hierarchical clustering method with B-spline knots to group USV functions by shape.
• Applied multiple knots for defining breakpoints in discontinuous USV calls.

Main Results:

• Successfully reduced noise and accurately detected USV calls.
• Developed a novel functional data analysis approach for USV characterization.
• Demonstrated effective hierarchical clustering of USV functions based on their shapes.
• Validated the proposed methods using laboratory mouse USV data.

Conclusions:

• The improved methods offer a significant advancement in the automatic analysis of mouse ultrasonic vocalizations.
• This approach facilitates more detailed and objective studies of rodent communication.
• The functional data analysis and clustering techniques provide a powerful tool for bioacoustic research.