identifying music via waveforms
-
Hello, I'm not sure if this is the best place for this, but it's the first and only place I could think of. I have had an idea for a while and have done some quick research and was wondering what people thought about it and if anybody is interested in perhaps helping me do something with it. My idea has to do with identifying songs or audio files with some sort of waveform "thumbprint". An application that I thought of that could use this is an mp3 tagger for example. Text is a very unreliable method of matching for applications like this. Audio matching is the best option that I can think of. The thing is, is that waveform matching is not a new idea as it turns out. Here's a site I found that deals with this problem... http://www.cs.hmc.edu/clinic/projects/2001/auditude/description.html[^] The company auditude (www.auditude.com[^]) has some technology that creates audio "thumbprints" and they also have a product called ID3man (www.ID3man.com[^]) that does ID3 tag matching. I downloaded this product and it asks you a whole bunch of personal questions in order to use it which I find to be completely unnecessary and it also costs money to use it (which is not really all that bad, people need to make money some how). So the technology exists and auditude says that have 3 million+ audio signatures in their database to match against. But I'd like to create something similar that is a little more open and not so intrusive (no personal questions required for something so trivial). The CDDB vs. FreeDB thing is sort of what I'd like to do. I think auditude needs a little competition (if that's possible :)) So what do you people think? Don't bother? Great idea? I'm pretty clueless with audio, so that's where I'm pretty stuck. If anybody thinks they can help or just if you have comments then please let me know. I've been unemployed for 10 months now, so I've got lots of time on my hands. ;) Thanks, Andrew (noname123@shaw.ca)
-
Hello, I'm not sure if this is the best place for this, but it's the first and only place I could think of. I have had an idea for a while and have done some quick research and was wondering what people thought about it and if anybody is interested in perhaps helping me do something with it. My idea has to do with identifying songs or audio files with some sort of waveform "thumbprint". An application that I thought of that could use this is an mp3 tagger for example. Text is a very unreliable method of matching for applications like this. Audio matching is the best option that I can think of. The thing is, is that waveform matching is not a new idea as it turns out. Here's a site I found that deals with this problem... http://www.cs.hmc.edu/clinic/projects/2001/auditude/description.html[^] The company auditude (www.auditude.com[^]) has some technology that creates audio "thumbprints" and they also have a product called ID3man (www.ID3man.com[^]) that does ID3 tag matching. I downloaded this product and it asks you a whole bunch of personal questions in order to use it which I find to be completely unnecessary and it also costs money to use it (which is not really all that bad, people need to make money some how). So the technology exists and auditude says that have 3 million+ audio signatures in their database to match against. But I'd like to create something similar that is a little more open and not so intrusive (no personal questions required for something so trivial). The CDDB vs. FreeDB thing is sort of what I'd like to do. I think auditude needs a little competition (if that's possible :)) So what do you people think? Don't bother? Great idea? I'm pretty clueless with audio, so that's where I'm pretty stuck. If anybody thinks they can help or just if you have comments then please let me know. I've been unemployed for 10 months now, so I've got lots of time on my hands. ;) Thanks, Andrew (noname123@shaw.ca)
Hi, I am working on a personal project (not realeased yet) that creates catalogues of Disks and CDs. It then let you search in that catalogue , look for duplicates/uniques (using CRC32), etc. When i get everything ready i will release it (free). Then i will start working on something like the one you said. My first though is to read the first 30 sec of each MP3, normalize, smoth, and divide it to 3 basic friquencies (usung FFT). Let's say 5,10,15 Khz, calculate the total gain of those frequencies and store them in the catalogue as a Vector. After that, i have to find the distance of every Vector to each other. (actually i will use more friquences and use multidimensional vectors). The more close the vectors, the more similar the sounds. The user can choose the percentance of similarity. Please, let me know what you think... :) - - - - - - - - - - - - - - - - - - Memory leaks is the price we pay \0 01234567890123456789012345678901234
-
Hi, I am working on a personal project (not realeased yet) that creates catalogues of Disks and CDs. It then let you search in that catalogue , look for duplicates/uniques (using CRC32), etc. When i get everything ready i will release it (free). Then i will start working on something like the one you said. My first though is to read the first 30 sec of each MP3, normalize, smoth, and divide it to 3 basic friquencies (usung FFT). Let's say 5,10,15 Khz, calculate the total gain of those frequencies and store them in the catalogue as a Vector. After that, i have to find the distance of every Vector to each other. (actually i will use more friquences and use multidimensional vectors). The more close the vectors, the more similar the sounds. The user can choose the percentance of similarity. Please, let me know what you think... :) - - - - - - - - - - - - - - - - - - Memory leaks is the price we pay \0 01234567890123456789012345678901234
Ok, please mail me at nkast@fastnet.gr so we can talk about it. I am gonna send you the app i was talking about as soon as i get your mail address, to check it out. - - - - - - - - - - - - - - - - - - Memory leaks is the price we pay \0 01234567890123456789012345678901234