I would like to thank you and appreciate for all the effort that you are taking to make these videos. I am a graduate student who is passionate about signal processing and machine learning in audio and your content is the best so far I have found. Since there are very less resources these are all the more helpful
Thanks very much for this series. Your code example and patient explanation really helped me a lot for getting started in the field of AI audio signal processing
Wow, thank you for putting this together! Though I'm a musician, I don't have a background in audio signal processing and it's been a struggle to find a good compiled source of information for deep learning specific to audio domain problems.
Thank you for this series chief; I'd been looking at trying to build a detection algorithm using a fourier transform to produce spectrograms and the theory refresher has been super super useful Plus confirmation that I wasn't barking up the wrong tree into how to approach it is always nice
Your channel sounds exactly what I was looking for as an ASR researcher. Thank you for these informative videos. Do you have tutorials on Montreal Forced Alignment too?
Wow, I was watching the main series of Audio classification with Python you made. And when you explained the MFCCs passing as input to a CNN in your example you put this [100, 13, 1] as an input shape and you said the 100 came from the samples in the audio file (51200) / the hop length of 512 and I was about to ask where did the 51200 come from. Maybe with this series I'll be able to understand that and also the Mel Spectrogram which is something that I wanted to know about for the longest time from a non-so mathematical perspective, but rather a more code applicable one. Greetings to you and thank you very much
Thanks for these series. This is what I needed project that I had in my mind. Please don’t pull yourself back while going deep on theory :) I like to learn the theory along with practical examples.
Very excited for this Valerio! Your videos have helped me immensely at understanding audio and ML/DL in general. Quick question: Do you have any recommendations or resources or plan to contribute some discussion about adding environmental noise to audio? I would love to talk about data augmenting with noise / other sound files into existing signals with proper RMS and SNR as to not overpower the desired signal. I'd love to see how to add radio static / random noises into audio for purpose of simulating target radio speech data. Thanks for everything!
I don't think I'll cover data augmentation in this series. Unfortunately, I don't a have a reference for noise adding. However, you can find papers which tangentially talk about it. I'm thinking of papers which, for example, introduce denoising autoencoder architectures.
Can you please do a video on speech emotion recognition...I have already seen your speech recognition and that is not enough for emotion classification...but a good one
Hi Torsha, I may cover speech emotion recognition in the future. The closest problem I've tackled on this channel is genre classification n my DL for Audio series. You can re-use most of those concepts for your problem.
Hello, thank you for the tutorial. I am using java for developping android Apps and I am searching for an android library for sound comparison. I have tried "musicg" library but it gives bad results. Do you know any alternative ? thank you again
Do you know if AI could make polyphonic audio to MIDI faster and better? And what about the restoration of very old and mangled recordings? I really want to try to work on this if I can but it feels like a steep curve.
Hey Mr Hope you doing good ! Please Can you help me ? How Can we use speech recognition to detect falling in elderly people ? Just another question how to combine audio with image to implement fall detection ?? Thank you
I want to know about how to do Classification of heart sounds and Improve accuracy. Is the method used for heart sound classification the same as for audio classification?
We are trying to develop a robot , can we use this topic for understanding how to develop algorithm for the robot to listen to un natural sounds etc (not speech recognition)
Amazing content and so well explained!! how can i implement this knowledge in speaker verification? I already followed the previous videos and created a project which so far tries to identify the spoken user but despite the fact that im getting 1.00 accuracy the system still fails to identify correctly. Can anyone help?
Hi George, thank you! Serendipitously, a thread on speaker verification just came out on TheSoundofAI Slack community in the #advice channel. I suggest you go check that out and join the conversation ;)
This series will mainly be focused on audio DSP. You can check out my series on DL for Audio to get an idea of some of the architectures you can use for audio classification.
Thanks, brother, for your generousity! You deserve more than a subscription. I'm from a computer vision background your tutorial helped me a lot! Yet I can't able to join the soundofai community. Could you check the problem?
To simplify this more --> The model will extract feature and classify the audio in one of the three groups whether its [Good,Bad,AVG] i would like to know is making such kinda model is feasible ????
Can you try this link? ru-vid.com?event=video_description&v=iCwMQJnKk2c&q=https%3A%2F%2Fjoin.slack.com%2Ft%2Fthesoundofai%2Fshared_invite%2Fzt-f71npumr-anli6W4QCuZ8UCj2gLoBkw&redir_token=QUFFLUhqbjBEMUY2clB3UDhsZklLN29feEdaN1h0RXBFd3xBQ3Jtc0tsVHFzaXFtT2hvRC12eEYwRVdNSDFyOWJQb2J2ZGw3SnhIOXY0UmEwY3FKNmFXZ3R5REpvaFVHbXRhVVFndWJRWjdXYV9rdkdIRUljWTRXckhlV19vVVQwZ1VSSEJ2VFk2SFBTT0hTamh1SG5wUkRZbw%3D%3D
Thanks! Sorry, but I'm not aware of any such book. If you'd like to get a few ideas about general audio / music signal processing, you can check out this video where I suggest a few books on the topic ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-jb5F4PfuRrw.html Some of the ideas you'll learn there, can be re-used for time series data.
Hello, very interesting subject and video. I have an hugely important request to adress you. Is it possible to train an A.I so that it can compare a reference signal (input source signal) with an output signal (that is the modified result of the source signal) so it can calculate an corresponding signal for the output signal to be the same as the input signal? It may not be super clear but basically, the functional chain is: -an input signal that is a source of audio content -an output signal that is the audio content (input signal) but processed in a way that distord it temporally (out of phase for some part of the signal) and in amplitude. -a closed feedback loop that send the exact copy of the distorded output signal to a core processing unit that contain the ai. This feedback loop sits right in between the input signal and output signal so that the AI can align (by injecting a new signal to the input signal, a compensating signal) almost immediately the output signal with the input. I'm a student and hobbyist and your reply could really be a greatly appreciated and precious contribution to my project Thanks