Analyzing Changes in Voice Aging Effects in Audio Files

Hai everyone

"I have two audio files: one with my original voice (26 years old) and another one in which my voice has been aged using an app. I want to figure out what changes were made to the first audio file to make it sound like that. My goal is to apply these changes later to another audio file that is 20-30 years older. This way, the voice will sound older.

I've considered subtracting the FFT range of the two audios from each other and saving the changes at the end, but unfortunately, that didn't work. I have the same sentence in both audios, with the first being my original voice and the second being my aged voice. Do you have any ideas on how I can better identify and save the changes in the voice?"

Answer 1

[Shreshth]

Hey Mohamed,

I could see that using a FFT (Fast Fourier Transform) did not yield useful results for you in order to analyse the differences between your original voice recording and the aged version.

There are some alternatives like Spectral Analysis, Time domain Analysis, Voice Quality Analysis and Machine learning that can help you to achieve your objective in case you do not have a varied set of data sets to train a Machine learning model you could go for Spectral Analysis.

Instead of directly subtracting the FFT data, you can perform a more nuanced spectral analysis. This involves looking at the spectral characteristics such as formant frequencies, harmonics, and spectral envelope.

Kindly follow the steps below:

1. Formant Analysis: Older voices tend to have lower formant frequencies due to changes in the vocal tract. Analyse the formant frequencies of both recordings to see if there's a shift.
2. Harmonic Analysis: The aging voice can have a different harmonic structure due to changes in the vocal cords. Compare the harmonic-to-noise ratio between the two recordings.
3. Spectral Envelope: The shape of the spectral envelope can change with age, affecting the timbre of the voice. Use tools like LPC (Linear Predictive Coding) to analyse the envelope.

Using MATLAB you can perform the spectral analysis using Fourier Transform function. A detailed example can be found by clicking on the MathWorks documentation link below.

https://in.mathworks.com/help/matlab/math/basic-spectral-analysis.html#bve7skg-2

Here is an example that tells you how to view the spectrum of an audio signal in MATLAB.

https://in.mathworks.com/matlabcentral/answers/18357-viewing-spectrum-of-an-audio-signal-in-matlab

In addition to it you can also refer to Audio Toolbox that can be effectively used for Audio processing, speech analysis and acoustic measurement. Here is a reference link for it.

https://in.mathworks.com/help/audio/index.html?s_tid=CRUX_lftnav

Hope it helps,

Regards,

Shubham Shreshth