Hello,

I thonk that spectral mesure such as the spectral flatness measure could be of interest for dioscriminating between narrowband signal and broadband noise.

Sincerely,

Pascal

I think, the best way is to leverage or adopt some effective filtering technique, e.g., Butterworth Filtering, Kalman Filtering, Particle Filtering or the most recent is Grassmann Filtering. However., all this is based on your task and nature of data.

It depends what kind of signals you look for. If it is deterministic and you know its nature, that canbe explotited to increase the signal to noise ratio. An exampåle is recording impulse responses using sine sweeps or MLS signals. Slow sweeps may give better resolution and SN ratio.

If the noise is deterministic or time invariant, you are better off than if the noise does variate. However a signal is influenced by the background noise already from a 10 dB difference and up.

The human listening system is an example of a system that can operate under very low SN ratios because of the trained brain. Usually people rely on more information than the speech alone.

Thank you Pascal Scalart for your response.

Can you please provide me with the research paper or any other literature from where i can get more insight.

Thank You

Sukanya pandey

Thank You Muhammad Ali for your response.

Can you please provide me with the literature or any research paper which can give me some more insight on the methods you suggested.

Thanking You

Sukanya

Thank you Anders Buen for your response.

Actually signal is not deterministic, it is an unknown real time signal from which i have to discriminate between signal and noise.

Please provide me with the literature or research paper if you have any.

Thanks

Sukanya

The prior signal-to-noise ratio at each frequency is the basic characteristic for passive filtering algorithms.

Hello, Sukanya Pandey.

Features that compete with MFCCs are mel- or log-mel energies. The link below is related to a known Python library to compute sequencies of mel-energies of any 1-D signal, the called mel-spectrogram. In addition, there are others spectral based features with examples.

Best regards,

Caio

Depending on the application at hands. Looking at SNR at the beginning is somewhat helpful. To me, if the parameters you want to estimates are varying With time or so, KF and particle filtering could be good choices. But signal modeling must be known, anyway.

I hope this helps.

Thanku Caio Cesar Enside de Abreu

You have not attached the link (about the link you were talking about in your answer). Please attach the link that will be really helpful for me, as your answer seems to be helpful in my case.

Thanks

Sukanya Pandey