To access this work you must either be on the Smith College campus OR have valid Smith login credentials.

On Campus users: To access this work if you are on campus please Select the Download button.

Off Campus users: To access this work from off campus, please select the Off-Campus button and enter your Smith username and password when prompted.

Non-Smith users: You may request this item through Interlibrary Loan at your own library.

Publication Date

2025-5

First Advisor

Susan E. Voss

Second Advisor

Luca Capogna

Third Advisor

Nicholas J. Horton

Document Type

Honors Project

Degree Name

Bachelor of Arts

Department

Engineering

Keywords

Wideband acoustic immittance, acoustic leaks, deep learning, CNN, Grad-CAM

Abstract

Wideband Acoustic Immittance (WAI) is a non-invasive auditory diagnostic technique that evaluates human middle ear function. To effectively diagnose ear pathologies usingWAI, measurements from normal human ears must be well understood and carefully conducted without acoustic leaks. This thesis aims to develop a Convolutional Neural Network (CNN) to automatically identify measurements contaminated by acoustic leaks. The thesis begins by reviewing previous studies on the physical model of acoustic leak measurements, followed by a detailed description of the CNN development process. The experiment has two parts. The first part uses the NHANES WAI dataset, where acoustic leak measurements are labeled based on the rule from Groon 2015, and experiments are conducted using two types of input data dimensions. The performance of single-layer and double-layer convolutional models is analyzed. To improve performance, different frequency ranges are used as input, and Gradient-weighted Class Activation Mapping (Grad-CAM) is utilized to generate activation heatmaps to aid in understanding how the model makes its predictions. The second part uses the Smith WAI database, where synthetic acoustic leak data are generated based on high-quality measurements. Results demonstrate that CNNs are highly effective for classifying WAI acoustic leaks, achieving high accuracy. However, due to the absence of impedance magnitude in the NHANES dataset, there remains room for performance improvement. Future research directions to enhance model capabilities are suggested.

Rights

©2025 Jiayi Sun. Access limited to the Smith College community and other researchers while on campus. Smith College community members also may access from off-campus using a Smith College log-in. Other off-campus researchers may request a copy through Interlibrary Loan for personal use.

Language

English

Comments

128 pages : color illustrations, Includes bibliographical references (pages 102-105).

Download

Smith Only:
Off Campus Download

Share

COinS