MATLAB code for language identification using audio signal

what is th ematlab code for identifying language using its audio signals.

Answer 1

[Dheeraj Singh]

For any identifying language you will need huge dataset of the words in that language if you want to use any Machine learning techniques. Also, the letters should be in English letters as most of the speech to text converter give output as English letters.

For converting audio to text, you can use the following file exchange link:

https://in.mathworks.com/matlabcentral/fileexchange/65266-speech2text?s_tid=prof_contriblnk

After getting text you can refer to the following link for text classification using Deep Learning:

https://in.mathworks.com/help/textanalytics/ug/classify-text-data-using-deep-learning.html

For more on speech and text classification refer to the following link:

https://in.mathworks.com/discovery/natural-language-processing.html

[Walter Roberson]

I wonder if the question is about identifying which language was being spoken, rather than about identifying what was being spoken?

[Dheeraj Singh]

I'm not sure if we can identify the language without extracting what is being spoken.

[Walter Roberson]

Sufficiently different languages have different statistics of sounds. It is possible for someone with experience to distinguish by subconsciously assessing the statistics even if they do not know the language being spoken.

If I hear a random language in passing, I can probably pick out the fact that it is German even though I have never studied German and it is not especially common where I am: I have encountered enough German in movies and so-on over the years to know what it sounds like. But I would have a hard time picking out whether it was Low German or High German or Swiss German.

It takes me a while to figure out whether someone is peaking Swedish or Danish.

Italian has been pretty common in media over the years, even if only in stereotypes; I can easily pick out out from (say) Spanish. But I might "hear" Castilian Spanish as if it were a different language than Latin American Spanish.

I cannot distinguish the Russian-like languages by ear with the (lack of) experience I have. But I can tell Ukranian from Russian without understanding either.

[Dheeraj Singh]

Yeah sure, we can do it using only the audio signal if we only want the language being spoken and not what is being spoken.

In this case we can use the Audio toolbox for feature extraction of the audio signal and then classify them using different Machine Learning techniques.