Wiener Noise Removal
When annoying background noise is drowning out the conversation, we can provide adaptive noise removal algorithms. In order to be sucessfull, the desired signal must not be WSS (wide sense stationary) - basically meaning it must be complex, and vary alot, and preferably have adequate times where the desired signal energy is zero. The noise component needs to be WSS, that is, consistent spectrum background noise.
In such cases it is possible to build something that adapts to the consist noise present, builds a spectral representation of it - and subtracts it from the overall signal + noise, leaving just the desired signal.
Typical applications: Removal of fan noise near a microphone, removal of some repetitive noises, such as road noise, or engine drone. DSPWidgets DSP consulting can provide a royalty free solution custom made that rivals the cost of licensing for a year.
When annoying background noise is drowning out the conversation, we can provide adaptive noise removal algorithms. In order to be sucessfull, the desired signal must not be WSS (wide sense stationary) - basically meaning it must be complex, and vary alot, and preferably have adequate times where the desired signal energy is zero. The noise component needs to be WSS, that is, consistent spectrum background noise.
In such cases it is possible to build something that adapts to the consist noise present, builds a spectral representation of it - and subtracts it from the overall signal + noise, leaving just the desired signal.
Typical applications: Removal of fan noise near a microphone, removal of some repetitive noises, such as road noise, or engine drone. DSPWidgets DSP consulting can provide a royalty free solution custom made that rivals the cost of licensing for a year.
Time Domain Echo Cancellation
Sometimes its possible to isolate the part of the signal you consider noise with a known copy. This is the case when you are playing music where a microphone is present. Both speech and the music come through the mic, but the good news is, your platform is rendering the music, so you have a copy of it. You would think that you can simply subtract the music reference from the mic signal, but you don't know the intensity of what reaches the mic, or the phase delay. Further, the music bounces off the walls and other objects present creating multiple reflections.
Echo cancellation is the process of of finding a correlation component of the noise and removing it. This component may have more than one time shifted copy.
Advanced techniques include breaking the spectrum up in bands, which also reduces the sampling rate needed for the sub bands, then processing the echo cancellation on each band, then reassembling the result, bumping the rate back to normal. This improves the echo removal or ERLE (Echo Return Loss Enhancement).
Simple and reliable solutions exist, some even in the public domain, but porting for performance and measuring the results is difficult. Using a DSP consultant is a better choice and can save you needless royalty payments.
Sometimes its possible to isolate the part of the signal you consider noise with a known copy. This is the case when you are playing music where a microphone is present. Both speech and the music come through the mic, but the good news is, your platform is rendering the music, so you have a copy of it. You would think that you can simply subtract the music reference from the mic signal, but you don't know the intensity of what reaches the mic, or the phase delay. Further, the music bounces off the walls and other objects present creating multiple reflections.
Echo cancellation is the process of of finding a correlation component of the noise and removing it. This component may have more than one time shifted copy.
Advanced techniques include breaking the spectrum up in bands, which also reduces the sampling rate needed for the sub bands, then processing the echo cancellation on each band, then reassembling the result, bumping the rate back to normal. This improves the echo removal or ERLE (Echo Return Loss Enhancement).
Simple and reliable solutions exist, some even in the public domain, but porting for performance and measuring the results is difficult. Using a DSP consultant is a better choice and can save you needless royalty payments.
Homomorphic Filtering
What is that? Well in some cases, when signal present in an environment is captured by a transducer, the transducer or the environment causes a convolution of the signal to occur. This is because of the physical characteristics of the room, or device. The convolution can cause multiple copies of the signal (such as acoustic room echo) - or act as an undesired spectral filters, because of poor transducer quality.
When the convolution tends to be stationary with time, but the signal not so much, it is possible to deduce the convolution impulse response from only the distorted signal. This is in fact an adaptive filter, and can be used to undo some of the damage. Applications include enhancing old recordings to removal of conference room echo (that I am in a small room with hard walls effect).
We are working on a solution today, that can improve your audio processing pipeline quality - a piece of the holistic solution provided by a DSP consulting service, not just another DSP contractor.
What is that? Well in some cases, when signal present in an environment is captured by a transducer, the transducer or the environment causes a convolution of the signal to occur. This is because of the physical characteristics of the room, or device. The convolution can cause multiple copies of the signal (such as acoustic room echo) - or act as an undesired spectral filters, because of poor transducer quality.
When the convolution tends to be stationary with time, but the signal not so much, it is possible to deduce the convolution impulse response from only the distorted signal. This is in fact an adaptive filter, and can be used to undo some of the damage. Applications include enhancing old recordings to removal of conference room echo (that I am in a small room with hard walls effect).
We are working on a solution today, that can improve your audio processing pipeline quality - a piece of the holistic solution provided by a DSP consulting service, not just another DSP contractor.