Spectral Smoothing for Frequency-domain Blind Source Separation
نویسندگان
چکیده
This paper describes the circularity problem of frequencydomain blind source separation (BSS), and presents a new method for solving it. Frequency-domain BSS performs independent component analysis (ICA) in each frequency bin. It is more efficient than time-domain BSS where ICA is applied to convolutive mixtures. However, frequency-domain BSS has two problems. The first is the permutation problem, for which we have recently proposed a method. It provides a robust and precise solution for the permutation problem and reveals the influence of the second problem, namely the circularity problem. Our solution for this second problem is based on spectral smoothing by windowing. However, the direct application of windowing changes the frequency responses for separation obtained by ICA and causes an error. Therefore, we adjust the frequency responses before windowing so that the error is minimized. The effectiveness of the method is shown by experimental results for the separation of up to four sources.
منابع مشابه
Blind Spectral-GMM Estimation for Underdetermined Instantaneous Audio Source Separation
The underdetermined blind audio source separation problem is often addressed in the time-frequency domain by assuming that each time-frequency point is an independently distributed random variable. Other approaches which are not blind assume a more structured model, like the Spectral Gaussian Mixture Models (Spectral-GMMs), thus exploiting statistical diversity of audio sources in the separatio...
متن کاملModulation domain blind source separation for noisy speech mixture
In this paper, we propose a noise-robust blind speech separation (BSS) method by using two microphones. We first use modulation domain real and imaginary spectral subtraction (MRISS) to enhance both magnitude and phase spectra of the speech mixture inputs. We then estimate the direction of arrivals (DOAs) of the speech sources and perform time-acoustic-modulation frequency masking to recover th...
متن کاملA Time-Frequency Domain Underdetermined Blind Source Separation Algorithm for MIMO Radar Signals
This paper considers the underdetermined blind separation of multiple input multiple output (MIMO) radar signals that are insufficiently sparse in both time and frequency domains under noisy conditions, while traditional algorithms are usually applied in the ideal sparse environment. An effective separation method based on single source point (SSP) identification and time-frequency smoothed l0 ...
متن کاملSéparation de sources par lissage cepstral des masques binaires (Source separation by cepstral smoothing of binary masks) [in French]
Source separation by cepstral smoothing of binary masks In this paper, we propose a separation system of speech signals from two convolutive mixtures. The suggested system is based on the combination of blind source separation technique with a time-frequency masking procedure, followed by a smoothing cepstral. Indeed, after separation of signal sources, the estimated binary masks undergo a ceps...
متن کاملPerformance of Quadratic Time-Frequency Distributions in Blind Source Separation
Blind source separation based on time-frequency distributions (TFD’s) allows the separation of Gaussian sources with identical spectral shape but different t-f localization properties. As TFD’s spread noise power while localizing the source energy in the t-f domain, the time-frequency BSS is characterized by robustness of separation and improved overall performance. However, quadratic TFD’s, wh...
متن کامل