Hi everyone i am doing my project in speech processing which involves removal of musical noise after spectral subtraction for removal of noise in speech signal . i have two problems. 1)i dont know how "musical noise" sounds like.if anyone could post here some link of audio files that do have musical noise then that would be great 2)spectral subtraction requires estimation of noise.can anyone give some link or reference for a simple algorithm that could be coded in matlab for estimation of noise in noisy speech signal. please note that this project is using single microphone so please do not introduce methods that might involve microphone arrays. Thankyou Khurram Hameed
regarding musical noise
Started by ●July 16, 2008
Reply by ●July 16, 20082008-07-16
khurram712 wrote:> Hi everyone > > i am doing my project in speech processing which involves removal of > musical noise after spectral subtraction for removal of noise in speech > signal .Hello Khurram Your approach is backwards. Instead of fixing the wrong solution you should find the right solution. The spectral subtraction can be modified so that it does not produce "musical noise". There is nothing more that I may say on this topic, but sometimes knowing that a solution exists helps in finding it. Regards, Andor
Reply by ●July 16, 20082008-07-16
khurram712 wrote:> Hi everyone > > i am doing my project in speech processing which involves removal of > musical noise after spectral subtraction for removal of noise in speech > signal .As noted by Andor, the musical noise is the artifact of the crude noise reduction performed at the first step. Not much can be done about the musical noise afterwards; you have to improve the initial algorithm.> i have two problems. > > 1)i dont know how "musical noise" sounds like.if anyone could post here > some link of audio files that do have musical noise then that would be > greatIt depends. Musical noise sounds kinda "watery" to me.> 2)spectral subtraction requires estimation of noise.can anyone give some > link or reference for a simple algorithm that could be coded in matlab for > estimation of noise in noisy speech signal.For the purpose of the student project, you can estimate the noise in the pauses of speech.> please note that this project is using single microphone so please do not > introduce methods that might involve microphone arrays.Alas. Not much of improvement can be done with a single mike. Vladimir Vassilevsky DSP and Mixed Signal Design Consultant http://www.abvolt.com
Reply by ●July 16, 20082008-07-16
On Jul 16, 12:32�am, "khurram712" <khurram...@hotmail.com> wrote:> Hi everyone > > i am doing my project in speech processing which involves removal of > musical noise after spectral subtraction for removal of noise in speech > signal . > > i have two problems. > > 1)i dont know how "musical noise" sounds like.if anyone could post here > some link of audio files that do have musical noise then that would be > great > > 2)spectral subtraction requires estimation of noise.can anyone give some > link or reference for a simple algorithm that could be coded in matlab for > estimation of noise in noisy speech signal. > > please note that this project is using single microphone so please do not > introduce �methods that might involve microphone arrays. > > Thankyou > > Khurram HameedWow Khurram, You decided on a topic for a project and it seems this is the first place you go! Whatever happened to doing a little research and asking about what you don't understand after you put some time in trying to learn? My understanding is that this is YOUR project. You are supposed to LEARN something from it, not simply have something to turn in. Take a little time and think. Figure out where the musical noise comes from, it is explained in some of the articles on this topic. Do a simulation in MATLAB of the processing for spectral subtraction, it is very few lines of code. If you want to hear the "musical" noise, a good estimate of the actual noise is not required. What galls me about how you are asking for help is that you sound so much like some of the working engineers who come here to avoid work and understanding. They will probably produce lousy systems, because they don't know enough to know if their system works properly. Hopefully you won't be one of them after you graduate. Dirk
Reply by ●July 16, 20082008-07-16
dbell wrote:> What galls me about how you are asking for help is that you sound so > much like some of the working engineers who come here to avoid work > and understanding. They will probably produce lousy systems, because > they don't know enough to know if their system works properly. > Hopefully you won't be one of them after you graduate.Dirk, In the latest "Embedded System Programming", Jack Ganssle (BTW he is about the only one worth reading in that magazine) stated that the good engineer is the "resource hunter". I.e. take the modules from MatLab, pull some code from the web, put it in one pile and ship. If there is no canned solution, drop the project and do something else. Being a geek doesn't pay neither for you nor for the company. As simple as that. With this sort of attitude, I am just wondering why the airplanes are not falling on my head at every five minutes.> > DirkVLV
Reply by ●July 16, 20082008-07-16
On Jul 16, 12:01�pm, Vladimir Vassilevsky <antispam_bo...@hotmail.com> wrote:> dbell wrote: > > What galls me about how you are asking for help is that you sound so > > much like some of the working engineers who come here to avoid work > > and understanding. �They will probably produce lousy systems, because > > they don't know enough to know if their system works properly. > > Hopefully you won't be one of them after you graduate. > > Dirk, > > In the latest "Embedded System Programming", Jack Ganssle (BTW he is > about the only one worth reading in that magazine) stated that the good > engineer is the "resource hunter". I.e. take the modules from MatLab, > pull some code from the web, put it in one pile and ship. If there is no > canned solution, drop the project and do something else. Being a geek > doesn't pay neither for you nor for the company. As simple as that. > With this sort of attitude, I am just wondering why the airplanes are > not falling on my head at every five minutes. > > > > > Dirk > > VLVHi Vlad, That attitude is a shame. Is that really an engineer? I am not opposed to using what already exists, but I at least need to be convinced that it does what it is supposed to. Which means that I first need to understand what it is supposed to do. I just don't have the blind faith necessary to take something and assume it works. I guess "Blind Faith 101" was a class they started after I graduated. I'll start watching the sky for the planes. Dirk
Reply by ●July 16, 20082008-07-16
On 16 Jul, 18:01, Vladimir Vassilevsky <antispam_bo...@hotmail.com> wrote:> dbell wrote: > > What galls me about how you are asking for help is that you sound so > > much like some of the working engineers who come here to avoid work > > and understanding. �They will probably produce lousy systems, because > > they don't know enough to know if their system works properly. > > Hopefully you won't be one of them after you graduate. > > Dirk, > > In the latest "Embedded System Programming", Jack Ganssle (BTW he is > about the only one worth reading in that magazine) stated that the good > engineer is the "resource hunter". I.e. take the modules from MatLab, > pull some code from the web, put it in one pile and ship. If there is no > canned solution, drop the project and do something else. Being a geek > doesn't pay neither for you nor for the company. As simple as that. > With this sort of attitude, I am just wondering why the airplanes are > not falling on my head at every five minutes.If not planes, so buildings. Yesterday a bridge, still under construction, collapsed in Stockholm, Sweden, with one fatality and two injured still uncertain to survive (Norwegian text only, as this Norwegian newspaper only publishes domestic news in English): http://www.aftenposten.no/nyheter/uriks/article2541888.ece Five minutes ago I saw on TV that the the collapse happened as some beams were put in place as part of the construction, and that the likely cause for the collapse was that the structure fell apart under the added load. Which, if true, means that the bridge was not designed even to support its own weight. Rune
Reply by ●July 16, 20082008-07-16
On Jul 16, 12:01�pm, Vladimir Vassilevsky <antispam_bo...@hotmail.com> wrote:> In the latest "Embedded System Programming", Jack Ganssle (BTW he is > about the only one worth reading in that magazine) stated that the good > engineer is the "resource hunter". I.e. take the modules from MatLab, > pull some code from the web, put it in one pile and ship. If there is no > canned solution, drop the project and do something else.Many years ago I worked for a manager who was trained as a computer scientist. He and I were constantly frustrated with each other because he would never give me enough information to do my job and I would therefore never live up to his expectations. One day when he and I were discussing this problem, I said, "Let's assume that you've just given me an assignment to design something. It doesn't really matter what, but let's just say that it's a dynamic range compressor. After I leave your office, you become invisible and secretly follow me back to my desk. What would you expect to see me do?" He answered, "I'd expect to see you pull your DSP book from your bookcase, open it to the chapter on dynamic range compressors, and start coding." At that point I knew for certain that there was no way he and I would ever see eye-to-eye. Greg
Reply by ●July 16, 20082008-07-16
Rune Allnor wrote:> On 16 Jul, 18:01, Vladimir Vassilevsky <antispam_bo...@hotmail.com> > wrote: >> dbell wrote: >>> What galls me about how you are asking for help is that you sound so >>> much like some of the working engineers who come here to avoid work >>> and understanding. They will probably produce lousy systems, because >>> they don't know enough to know if their system works properly. >>> Hopefully you won't be one of them after you graduate. >> Dirk, >> >> In the latest "Embedded System Programming", Jack Ganssle (BTW he is >> about the only one worth reading in that magazine) stated that the good >> engineer is the "resource hunter". I.e. take the modules from MatLab, >> pull some code from the web, put it in one pile and ship. If there is no >> canned solution, drop the project and do something else. Being a geek >> doesn't pay neither for you nor for the company. As simple as that. >> With this sort of attitude, I am just wondering why the airplanes are >> not falling on my head at every five minutes. > > If not planes, so buildings. Yesterday a bridge, still under > construction, collapsed in Stockholm, Sweden, with one fatality > and two injured still uncertain to survive (Norwegian text only, > as this Norwegian newspaper only publishes domestic news in English): > > http://www.aftenposten.no/nyheter/uriks/article2541888.ece > > Five minutes ago I saw on TV that the the collapse happened > as some beams were put in place as part of the construction, > and that the likely cause for the collapse was that the structure > fell apart under the added load. Which, if true, means that the > bridge was not designed even to support its own weight.Concrete construction? Concrete gets stronger as it ages, but the strength at 30 days is used in load calculations. The seven-day strength is also specified; that's important for timing the form removal and for setting forms for the next level. Maybe the 32-day strength would have been adequate, but the builder didn't wait long enough. Monday in New York, a building inspectors' car was demolished by a bucket of concrete weighing over a ton that fell about 30 stories from a building being renovated. Two inspectors who had just finished their inspection tour were about to get into the car, but were uninjured. It would be interesting to know what they had inspected. Jerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������
Reply by ●July 23, 20082008-07-23
In article <htSdnXbCo-p25ODVnZ2dnUVZ_sDinZ2d@giganews.com>, khurram712@hotmail.com says...> > >Hi everyone > >i am doing my project in speech processing which involves removal of >musical noise after spectral subtraction for removal of noise in speech >signal . > >i have two problems. > >1)i dont know how "musical noise" sounds like.if anyone could post here >some link of audio files that do have musical noise then that would be >great > >2)spectral subtraction requires estimation of noise.can anyone give some >link or reference for a simple algorithm that could be coded in matlab for >estimation of noise in noisy speech signal. > >please note that this project is using single microphone so please do not >introduce methods that might involve microphone arrays.Here is a reference [Wolfe & Godsill: "The Application of Psychoacoustic Criteria to the Restoration of Musical Recordings," AES Preprint 5150, presented at the 108th Convention, 2000 February We consider the restoration of musical recordings degraded by additive broadband noise such as tape hiss, where maintaining fidelity of the underlying musical signal is of primary importance. By integrating psychoacoustics into a restoration process based on perceptual optimality criteria, we take advantage of human auditory perception to optimize the compromise between noise reduction and restored signal fidelity.]
