About Me
Welcome! I'm Noah. I am a PhD student in the Phonetics Lab at UCSD and a Research Scientist Intern at Meta Reality Labs.
My research revolves around evaluating the performance of different voice AI models, focusing on zero-shot speaker adaptation using speech acoustics and dynamic signal processing. At Meta, I am focusing on creating audio quality evaluation pipelines for Conversational Focus, which is a hearing enhancement tool found on Meta glasses.
Speech is one of the most complex things we do as humans. When someone speaks, regardless of what they are saying, the listener perceives supra-linguistic information about the speaker. This can include biological information such as size, weight, and age of the speaker, or social information like their race/ethnicity, gender, and even socioeconomic status. The listener can also perceive information about the emotional state of the speaker; whether they sound happy, sad, angry, bored, etc. All the while, everyone has their own unique voice and sounds different from one another.
Modern speech technology has advanced to the point of trying to imitate many of these qualities in synthetic speech from very little input data. However, these things are difficult to assess. That's where my research comes in: With advances in speech technology, comes the need for advanced evaluation techniques and better training data.