Dual Square Root Unscented Kalman Filter based Single Channel Blind Source Separation Methodology

Dutt, Rashi and Acharyya, Amit and Sheikh, Israr (2022) Dual Square Root Unscented Kalman Filter based Single Channel Blind Source Separation Methodology. In: 2022 IEEE International Symposium on Circuits and Systems, ISCAS 2022, 27 May-1 June 2022, Austin.

[img] Text
Proceedings_IEEE_International_5.pdf - Published Version
Restricted to Registered users only

Download (1MB) | Request a copy

Abstract

Single channel Blind Source Separation (SCBSS) is a challenging problem for several real-world practical applications. The existing SCBSS methodologies depend upon the properties of the sources present in the mixture and hence do not remain truly blind. Also, the solutions are found to be suboptimal and limited in application. In this paper, we present an SCBSS methodology using a state-parameter estimation approach to eliminate the constraints on the source signals such as statistical independence and frequency disjoint spectra. A Dual Square Root Unscented Kalman Filter (D-SRUKF) estimator has been proposed, which demonstrates higher numerical accuracy and improved stability compared to the widely used Dual Extended Kalman Filter (D-EKF). Simulations have been performed for separating mixed signals with overlapping spectra such as speech and biomedical signals. The proposed methodology demonstrates higher Signal-to-Interference Ratio (SIR) and Signal-to-Distortion Ratio (SDR) when the current methodologies even fail to separate the sources. The results also show that the proposed DSRUKF SCBSS is 15% more accurate than the state-of-the-art D-EKF SCBSS and has higher stability owing to the square root formulation of D-SRUKF source estimator. © 2022 IEEE.

[error in script]
IITH Creators:
IITH CreatorsORCiD
Acharyya, Amithttp://orcid.org/0000-0002-5636-0676
Item Type: Conference or Workshop Item (Paper)
Additional Information: The authors acknowledge the Intel India Research Fellowship - 2020 for supporting the project. Authors would also like to acknowledge the support extended by the Defence Research and Development Organisation (DRDO), Ministry of Defence, Government of India.
Uncontrolled Keywords: Dual Square Root Unscented Kalman Filter; Single channel Blind Source Separation; State-Parameter Estimation
Subjects: Electrical Engineering
Divisions: Department of Electrical Engineering
Depositing User: Ms Palak Jain
Date Deposited: 22 May 2023 09:30
Last Modified: 22 May 2023 09:30
URI: http://raiith.iith.ac.in/id/eprint/11472
Publisher URL: https://doi.org/10.1109/ISCAS48785.2022.9937287
Related URLs:

Actions (login required)

View Item View Item
Statistics for RAIITH ePrint 11472 Statistics for this ePrint Item