Project Summary

The purpose of this Senior Capstone Project is to design and build five independent models of the human vocal tract, one each for five distinct and recognizable vowel sounds. The models will provide a visual and auditory educational aid to of students and instructors at varying levels of academia with the goal of increasing one’s understanding of speech acoustics. The models will be translucent cylinders with a smooth exterior and a visible curved interior contour modeled after the shape of the vocal tract for each vowel. Each model have been built in Solidworks using the simple sketch, extrude, and loft cut features. They will be capable of generating accurate phonemes or speech samples for the given vowel sound.

Tests for accuracy will occur at two stages in the project: before construction and after. The first round of testing, to be done over the winter break, consists of taking the completed Solidworks file and testing them with a Matlab code called VTcalcs provided by Professor Hanson from Satrajit S. Ghosh, a published research scientist at the McGovern Institute for Brain Research at MIT and an assistant professor Department of Otology and Laryngology at Harvard Medical School. Once the models are deemed accurate in comparison to commonly accepted transfer functions and format frequencies, the Solidworks files will be sent to one of a number of companies to be 3D printed or molded in acrylic resin. After the physical models are received, they will be tested a second time using a recorded audio signal generated by each model and a computer program that will produce a spectrogram from which the formant frequencies and can be measured and compared. Additionally, an informal listening test with a handful of volunteers will be conducted to measure practical accuracy and usage of the models. In order for the project to be classified as successful, the models will have a visible interior contour in the shape of the vocal tract and be able to produce an accurate speech sample for each vowel that can be easily identified by the user.