Students: Yulin Zhou
Project: Discrete Speech Limited Vocab Recognition with Application in Verbal equation Writing
Poster: Vertical (PDF) | Horizontal (PDF)
Institution: Lafayette College
Major: Electrical and Computer Engineering
Advisor: Ismail Jouny
Abstract
In this modern age, speech recognition plays a critical role in hands-free driving, security, home integration systems, and disability assistance. While current speech recognition software has very holistic solutions and is very well-developed, many real usages of speech recognition technology are more situation-specific. This paper aims to apply a limited library of vocabulary, derived from mathematical equations, and test the accuracy of different machine learning algorithms such that various performances can be compared. This experiment reveals that limited vocab speech recognition can achieve its best performance using the Convolutional Neural Network(CNN). The CNN takes in the data processed by dynamic time warping, combined with an enlarged dataset by adding white noise to the original audio inputs to obtain the Mel-frequency Cepstral coefficients for further analysis.
About Yulin Zhou
Yulin Zhou, senior Electrical and Computer Engineering major and a Computer Science minor at Lafayette College from Shanghai, China. Working on a research project with Professor Ismail Jouny which focuses on exploring various algorithms in machine learning and signal processing. Will pursue a master degree in Carnegie Mellon University in ECE after graduation. Having voice lessons since the beginning of this semester and very much enjoy singing.