hello sir,, the tutorial is easy to understand because the explanation is very clear. but I found a problem in this section "'ls' is not recognized as an internal or external command, operable program or batch file." please help why this part cannot be recognized. what should I install
Thanks for your very interesting video! A question: but if I have to align 2 tracks with different BPM (one 108 and the other 120 bpm) what can I do? do I raise the first or lower the second? or do I take them both at 119 bpm? but will the beat grid be constant for the 2 files?
I believe so, feature extraction is based on the input content so its best to match the BBP per channel for feature similarity. Thanks for the comment, appreciate it very much.
Which part of the tutorial do you think would be best for genre determination? I am trying to build a model/application that determines sub genres of Electronic music but might technical music knowledge is limited?
Thanks for the comment, appreciate it. You would have to use a combination of features to determine the genre, no single feature can be deterministic for the genre in the audio fragments. I hope the following article can help you more on this regard: www.clairvoyant.ai/blog/music-genre-classification-using-cnn
You would need to experiment with MEL Spectogram and other functions to get you started.. Please take a look at the following blog where human voice is recognize and sampled from a long minute audio. towardsdatascience.com/voice-classification-using-deep-learning-with-python-6eddb9580381
when i try the command plt.figure(figsize=(14, 5)) librosa.display.waveshow(music_array2, alpha=0.1) plt.vlines(beat_times, -1, 1, color='r') plt.ylim(-1, 1) it says : process_plot_var_args' object has no attribute 'prop_cycler' can you help mee?
Thanks for your comment, appreciate it. I believe that librosa can't read audio files from an in-memory buffer so unpacking the gz to temporary files is one option. Please read the following solution and hope it helps you: stackoverflow.com/questions/50202350/how-do-i-read-in-wav-files-in-gz
Hi sir, great video. I learned a lot. Is it possible to use the output from the feature extraction to create AI music composition? Or what else can I do with feature extraction besides creating genre prediction, transcription, and classification programs? I plan on doing feature extraction for my ML project and am looking to see what are the options. Thank you so much
Thanks for your comment. You can extract features and then embedded them together with the music notes and then play them using libraries i.e. tone.js. Other option is to apply extracted features with generative AI music creator libraries to create the music also.
Once option is to use the pydub library and then use make_chunks method to crate the same size chunks.. you can look at the documentation. Other option is create the manual chunks and then add padding.. The following link has full working code for you to follow: codereview.stackexchange.com/questions/221790/converts-audio-files-to-fixed-size-chunks-and-the-chunks-to-spectrogram-images Hope above help you, what you are looking for. Thanks for the comment.
I wish to quantize the data after applying DCT but I couldn't find a quantization table for audio data.Is there a way to do the quantization the right way?
thanks for the tutorial. how can i convert to time series data wrt to frames ex: (time_step, feature_dim) if my loaded audio data shape is (67032,) can i reshape it to (12,5586) #feature size and then repeat data with 3 time step to create (3,10,5586) I want to use this for lstm model
Sorry I did not understand your objective by reshaping your data? Do you want to feature engineer it? What is the logic/concept you would want to use with your feature engineering? If I know what you want, I can better answer your question. Thanks for the comment.
hello sir,, the tutorial is easy to understand because the explanation is very clear. but I found a problem in this section "'ls' is not recognized as an internal or external command,
operable program or batch file."
please help why this part cannot be recognized. what should I install
👍 Electroacoustic music composer thanks you ! 😀
Appreciate your comment, happy to hear that.
Thanks for your very interesting video!
A question: but if I have to align 2 tracks with different BPM (one 108 and the other 120 bpm) what can I do?
do I raise the first or lower the second? or do I take them both at 119 bpm? but will the beat grid be constant for the 2 files?
I believe so, feature extraction is based on the input content so its best to match the BBP per channel for feature similarity. Thanks for the comment, appreciate it very much.
Amazing vedio thank you I learn more!! but I want to save the extracted feature to CSV file!! how can I do it? did you provide the souce code?
Which part of the tutorial do you think would be best for genre determination? I am trying to build a model/application that determines sub genres of Electronic music but might technical music knowledge is limited?
Thanks for the comment, appreciate it.
You would have to use a combination of features to determine the genre, no single feature can be deterministic for the genre in the audio fragments.
I hope the following article can help you more on this regard:
www.clairvoyant.ai/blog/music-genre-classification-using-cnn
thank you for your video, which method is best solution that i can process signal to recognise human voice like hello?
You would need to experiment with MEL Spectogram and other functions to get you started..
Please take a look at the following blog where human voice is recognize and sampled from a long minute audio.
towardsdatascience.com/voice-classification-using-deep-learning-with-python-6eddb9580381
thank you sir. the video was very helpful
Appreciate your comment, thank you so much.
Good Knowledge
when i try the command
plt.figure(figsize=(14, 5))
librosa.display.waveshow(music_array2, alpha=0.1)
plt.vlines(beat_times, -1, 1, color='r')
plt.ylim(-1, 1)
it says : process_plot_var_args' object has no attribute 'prop_cycler'
can you help mee?
Very good 👍👍
Hey the dataset i have is with gz format .how do i work with that ?could you give me pointers
Thanks for your comment, appreciate it.
I believe that librosa can't read audio files from an in-memory buffer so unpacking the gz to temporary files is one option.
Please read the following solution and hope it helps you:
stackoverflow.com/questions/50202350/how-do-i-read-in-wav-files-in-gz
Hi sir, great video. I learned a lot. Is it possible to use the output from the feature extraction to create AI music composition? Or what else can I do with feature extraction besides creating genre prediction, transcription, and classification programs? I plan on doing feature extraction for my ML project and am looking to see what are the options. Thank you so much
Thanks for your comment. You can extract features and then embedded them together with the music notes and then play them using libraries i.e. tone.js.
Other option is to apply extracted features with generative AI music creator libraries to create the music also.
How can I divide an audio data into several equal sized chunks with padding? ( I'm dividing the data into chunks to apply DCT on every chunk)
Once option is to use the pydub library and then use make_chunks method to crate the same size chunks.. you can look at the documentation.
Other option is create the manual chunks and then add padding.. The following link has full working code for you to follow:
codereview.stackexchange.com/questions/221790/converts-audio-files-to-fixed-size-chunks-and-the-chunks-to-spectrogram-images
Hope above help you, what you are looking for. Thanks for the comment.
Thank you so much for your response.I really appreciate it.
I wish to quantize the data after applying DCT but I couldn't find a quantization table for audio data.Is there a way to do the quantization the right way?
this video is enough for audio processing?
Thanks for the comment, appreciate it. Yes, you can follow the full tutorial and it will get you from start to finish.
thanks for the tutorial.
how can i convert to time series data wrt to frames
ex: (time_step, feature_dim)
if my loaded audio data shape is (67032,)
can i reshape it to (12,5586) #feature size
and then repeat data with 3 time step to create (3,10,5586)
I want to use this for lstm model
Sorry I did not understand your objective by reshaping your data? Do you want to feature engineer it? What is the logic/concept you would want to use with your feature engineering?
If I know what you want, I can better answer your question. Thanks for the comment.