Task: The human phenotype vector

[satra]

Description

The idea is to start defining and populating different aspects of the human phenotype estimated from a voice recording.

Tasks

• Define buckets (e.g. physiology, voice quality, rate, emotion, ...)
• Define features in each bucket
• Prioritize features
• Implement features
• Validate features

Freeform Notes

ping @900miles and @wilke0818

[wilke0818]

It is unclear to me what this even means. perhaps we can discuss at a later point.

[fabiocat93]

It is unclear to me what this even means. perhaps we can discuss at a later point.

@wilke0818 this should be one of the multiple project options we have discussed. I believe the goal here is to establish adn implement a set of descriptors for voice and speech. These descriptors should cover various levels of abstraction and interpretability. For example, they might include low-level acoustic features as well as high-level info like emotions or transcripts. The main motivation is that we want to shift from a class-based description of voice to a dimensional space that captures the complexity of voice and speech.

• input: audio file
• output: a dictionary (or maybe an array given the title of the issue) describing the speaker + their behaviors in the clip