Thanks for your answer Pichat. But still I am not entirely sure on how to generate this in Matlab.

As I am not that deep in details about special matlab functions, I would like to point out the main characteristics of the Bark scale again.

The Bark scale is based on the frequency resolution of human hearing, based on critical bands of our perception. Therefore, it is more or less logarithmic at higher frequencies, more linearly spaced at the lower end (and thus not identical to 1/3 octaves, etc.).

If you want to do that kind of analysis based on an FFF, you need to align your frequeny bins to the bark bands and sum them up.

However, be careful about the bandwith of your bins and also about potentially characteristics of filter banks in the cross-over ranges used in the literature if you want to compare results.

You might also try to implement a Bark based filter bank directly to better understand also hearing models.

where are you using this BandBarks in extracting Spectro. Additionally, you only understand if the results are making any sense or not by analysis your final result or application.

Tthe goal is song analysis and identification Krishna Chauhan

Analysis and identification?

Identification of genre or song?

If song, then merely lyrics will do, right?

In the case of the genre, acoustic properties are crucial.

Now back to your feature extraction, spectrogram using bark scale is not something very new; you extract the feature and train your model and see whether it is performing satisfactorily on your data or not.

In conclusion, you did not respond to my initial question there.

Regards

You can use lines 1 to 62 of the following code

https://github.com/Mak-Sim/modulation_spectra_in_CB/blob/master/MS_fft_analysis.m

to compute spectrogram in a Bark scale.

Maxim Vashkevich thank you very much for your help, I will take a look and try to implement it in my own code.