The human voice is such a fascinating tool, yet so complex. My project is heavily dependent on input signals based on human speech and so I considered doing some dedicated research on it. I was quite astonished by the results and was surprised to discover this world of phonetics and linguistics, which were so new to me.
I wanted to dive deeper and understand the ways the sounds in human speech differ and see just how varied they are. This brought me back to one of the assignments I did last semester for the Digital Signal Processing course. We were asked to record different vowels and analyze the distinctions between them. Formants were identified as the primary acoustic cues for vowel identification. The differences between the first two formant frequencies (F1 and F2) helped us distinguish each vowel.
For instance, in our exercise the vowel “U” only had a difference of 500 Hz, while a vowel like “E” had a difference of over 1600 Hz. The vocal tract acts as a filter, by altering the position of the tongue, jaw, and lips the speaker changes the shape of this resonating tube. This physical manipulation amplifies specific frequency bands (formants) while dampening others.
Apart from this, we also have consonants which possess their own unique characteristics. In English phonology, consonants can be split into 6 different categories based on their articulation. These include plosives, fricatives, affricates, nasals, approximants and laterals. These sounds are produced based on variations in airflow between the vocal tract and lips. Additionally, linguists have classified these consonants into 2 categories based on voicing: voiced (where the vocal cords vibrate when the sound is made) and voiceless (where the vocal cords do not vibrate, and the sound is shaped purely by breath).
While figuring out ways to test out these consonants with my prototype, I realized it would be cool to experiment with modern day beatboxing. I was inspired to do this while watching a WIRED video about the different levels of beatboxing. Popularized by the African American communities in the US during the mid 80s, it grew into a huge sensation and remains a staple in pop culture today. It would be a really cool way to experiment and analyze speech, as these artists use fricatives and plosives to recreate the rhythms similar to a drumkit. I am curious to see how this will turn out.
This also made me realize that I could add some sort of live audio detection tool for speech analysis. I am planning to search for a few Python libraries I could use to achieve this. This could be a very interesting add on to the project and would validate important information about the phonetic structure of the speech, providing valuable insights for singers, voice actors and other professionals working with the human voice.
