Unique Methods for Determining the Attenuation and Delay in Blind Source Separation Based on the Degenerate Unmixing Estimation Technique
The degenerate unmixing estimation technique (DUET), enables facile sound source separation through two signal mixtures. However, in numerous practical situations, identifying peaks in DUET histograms is difficult, which in turn prevents effective source separation. Specifically, sound sources with...
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2021-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/9540649/ |
| _version_ | 1811274089373892608 |
|---|---|
| author | Kuang-Yow Lian Jia-Hsin Lin |
| author_facet | Kuang-Yow Lian Jia-Hsin Lin |
| author_sort | Kuang-Yow Lian |
| collection | DOAJ |
| description | The degenerate unmixing estimation technique (DUET), enables facile sound source separation through two signal mixtures. However, in numerous practical situations, identifying peaks in DUET histograms is difficult, which in turn prevents effective source separation. Specifically, sound sources with short time percentages in mixtures are more unlikely to form peaks in a histogram. Besides, when the noise floor of mixtures is high, the height of the histogram peak may be reduced, which causes peak position bias and prevents effective source separation. In this study, unique methods are proposed to solve the aforementioned problems. A subhistogram is established for each time frame; thus, the original histogram is divided into a series of subhistograms. Two measurement indices are then proposed to identify the accurate attenuation and delay parameters for sound sources: maximum distribution, which is based on the infinity norm, and variance distribution, which is based on variances. The results verify that maximum distribution effectively highlights the peaks of instantaneous sound sources when the time percentage difference between sound sources is excessively high. In this scenario, the conventional histogram is unable to find the peaks. When a strong noise disturbance is present in mixtures, variance distribution can be employed to estimate peak positions with lower biases than those in the conventional histogram. In addition to more well-defined and intense peaks, the proposed method can also reduce the bias error by 71% when the SNR is 0 dB. Variance distribution is more robust against noise disturbance compared to conventional histograms and can be adopted in a wide range of applications. |
| first_indexed | 2024-04-12T23:12:35Z |
| format | Article |
| id | doaj.art-4e20ffaf59884b5aaaa6f987259e2912 |
| institution | Directory Open Access Journal |
| issn | 2169-3536 |
| language | English |
| last_indexed | 2024-04-12T23:12:35Z |
| publishDate | 2021-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Access |
| spelling | doaj.art-4e20ffaf59884b5aaaa6f987259e29122022-12-22T03:12:46ZengIEEEIEEE Access2169-35362021-01-01912946012947010.1109/ACCESS.2021.31134579540649Unique Methods for Determining the Attenuation and Delay in Blind Source Separation Based on the Degenerate Unmixing Estimation TechniqueKuang-Yow Lian0https://orcid.org/0000-0002-5692-9279Jia-Hsin Lin1https://orcid.org/0000-0002-0033-7565Department of Electrical Engineering, National Taipei University of Technology, Taipei, TaiwanDepartment of Electrical Engineering, National Taipei University of Technology, Taipei, TaiwanThe degenerate unmixing estimation technique (DUET), enables facile sound source separation through two signal mixtures. However, in numerous practical situations, identifying peaks in DUET histograms is difficult, which in turn prevents effective source separation. Specifically, sound sources with short time percentages in mixtures are more unlikely to form peaks in a histogram. Besides, when the noise floor of mixtures is high, the height of the histogram peak may be reduced, which causes peak position bias and prevents effective source separation. In this study, unique methods are proposed to solve the aforementioned problems. A subhistogram is established for each time frame; thus, the original histogram is divided into a series of subhistograms. Two measurement indices are then proposed to identify the accurate attenuation and delay parameters for sound sources: maximum distribution, which is based on the infinity norm, and variance distribution, which is based on variances. The results verify that maximum distribution effectively highlights the peaks of instantaneous sound sources when the time percentage difference between sound sources is excessively high. In this scenario, the conventional histogram is unable to find the peaks. When a strong noise disturbance is present in mixtures, variance distribution can be employed to estimate peak positions with lower biases than those in the conventional histogram. In addition to more well-defined and intense peaks, the proposed method can also reduce the bias error by 71% when the SNR is 0 dB. Variance distribution is more robust against noise disturbance compared to conventional histograms and can be adopted in a wide range of applications.https://ieeexplore.ieee.org/document/9540649/Attenuation and delay histogramblind source separationDUETinstantaneous sound sourcesW-disjoint orthogonality |
| spellingShingle | Kuang-Yow Lian Jia-Hsin Lin Unique Methods for Determining the Attenuation and Delay in Blind Source Separation Based on the Degenerate Unmixing Estimation Technique IEEE Access Attenuation and delay histogram blind source separation DUET instantaneous sound sources W-disjoint orthogonality |
| title | Unique Methods for Determining the Attenuation and Delay in Blind Source Separation Based on the Degenerate Unmixing Estimation Technique |
| title_full | Unique Methods for Determining the Attenuation and Delay in Blind Source Separation Based on the Degenerate Unmixing Estimation Technique |
| title_fullStr | Unique Methods for Determining the Attenuation and Delay in Blind Source Separation Based on the Degenerate Unmixing Estimation Technique |
| title_full_unstemmed | Unique Methods for Determining the Attenuation and Delay in Blind Source Separation Based on the Degenerate Unmixing Estimation Technique |
| title_short | Unique Methods for Determining the Attenuation and Delay in Blind Source Separation Based on the Degenerate Unmixing Estimation Technique |
| title_sort | unique methods for determining the attenuation and delay in blind source separation based on the degenerate unmixing estimation technique |
| topic | Attenuation and delay histogram blind source separation DUET instantaneous sound sources W-disjoint orthogonality |
| url | https://ieeexplore.ieee.org/document/9540649/ |
| work_keys_str_mv | AT kuangyowlian uniquemethodsfordeterminingtheattenuationanddelayinblindsourceseparationbasedonthedegenerateunmixingestimationtechnique AT jiahsinlin uniquemethodsfordeterminingtheattenuationanddelayinblindsourceseparationbasedonthedegenerateunmixingestimationtechnique |