habbes
commented
5 years ago Hey @LYoung520, thanks for checking this project out. This is an old quick project that I did as part of a job application, it's not something that I use, work on or maintain anymore. I shut down the server that did the inference and training, that's probably why you're not getting results.
However, the server source code is also available in this GitHub Repo.
So you can clone and host the server yourself and you can fork the android app and replace the BASE_URL in the MainActivity.java class with your own server url to test it out.
Also, note that the server does not attempt to predict the age, but broader age groups and gender (male child, male teen, male adult, male senior, female child, female teen, female adult, female senior).
The server made use of the pyAudioAnalysis library for the bulk of the training and inference tasks. You can check the library out to learn more about how it works. I have not kept up with the development and progress of the library and so I'm not sure what its current state is. I've also not dug deeper in audio analysis and inference since then so I can't help much. But I hear deep learning can give great results. So you could also give convolution and recurrent neural networks a try.
But if I get better insights and time, I might update the project or create a new one.
I am interested in predicting age from voice. I have tried your Android client but it seemed that I couldn't get any result from your server. I wonder if you can give me any clue about predicting age from voice such as a paper or a brief introduction of the training process.
Thanks a lot!