Application: Analyzing Voices#
Background#
One very informative, but complex signal is speech produced by the human voice. Speech contains direct information (sentences with words producing a story), and, when spoken, may also provide indirect information about the speaker. For example, Individuals with depression, for example, may speak (on average) less loud and slower.
Problem Statement | Significant Problem#
Your team is a group of auditory neuroscientists, hearing researchers and linguists analyzing voice data. The overall study aims to understand how differences in voices might be due to language, gender and cognitive state. Your task is to analyze the given voice data to identify patterns associated with gender.
Data and Resources | Same Problem#
The data consists of two sets of audio samples of a person speaking a coherent sentence. The two sets are from different people speaking. One is a female speaker, the other is a male speaker.
Task | Specific Choice#
Signal Analysis: Apply appropriate signal analysis techniques to transform the signal into something more readily interpretable.
Feature Identification: Identify the features that correlate with speaker gender.
Interpretation: Based on the analysis, teams should hypothesize how these signals reflect gender.
Reporting (Simultaneous Report)#
Each team will present their findings in a brief presentation (slides). Please use visual and auditory aids like graphs or charts to illustrate your findings.
Additional Considerations#
You will have to determine which feature is different in male and female voices. You are allowed to use information on the web to find out which feature that is.
you will have to do some coding (e.g. loading the signals into Matlab, doing signal analysis, making figures, etc). Again, you are allowed to use any tool (web, chatgpt, previous chapter) to do so.
Listen to some of the sounds.
The data set is limited and contains a confound. You are free to discuss this.
Follow-up#
After the presentation, we will see discuss how this application session went. In the next modules, we will see how elementary signals can be used to characterize (linear) systems.