Details
| Original language | English |
|---|---|
| Publication status | Published - 2017 |
| Event | DAGA 2017: 43. Jahrestagung für Akustik - Kiel, Germany Duration: 6 Mar 2017 → 9 Mar 2017 |
Conference
| Conference | DAGA 2017 |
|---|---|
| Country/Territory | Germany |
| City | Kiel |
| Period | 6 Mar 2017 → 9 Mar 2017 |
Abstract
Cite this
- Standard
- Harvard
- Apa
- Vancouver
- BibTeX
- RIS
2017. Paper presented at DAGA 2017, Kiel, Germany.
Research output: Contribution to conference › Paper › Research
}
TY - CONF
T1 - Analysis and Comparison of different Adaptive Filtering Algorithms for Fast Continuous HRTF Measurement
AU - Li, Song
AU - Peissig, Jürgen Karl
AU - Klinkert Correa, Camilo
PY - 2017
Y1 - 2017
N2 - Head-related transfer function (HRTF) is widely used for binaural sound reproduction over headphones. However, the acquisition of HRTFs using traditional measurement is usually a time-consuming task and can be only acquired at discrete directions. Recent work has shown, that the HRTF measurement can be speeded up and simplified via continuous acquisition by using an adaptive filter (identification of a time-variant system). With this method the traditional sampling and interpolation of many different positions can be avoided. There are many different adaptive filter types suitable for system identification, therefore this work analyses and compares various adaptive filter algorithms like Least-MeanSquares (LMS), Normalized-LMS (NLMS), RecursiveLeast-Squares (RLS), etc. The goal of this work is to achieve faster convergence speed and low steady-state mean squared error. Thus, the convergence and tracking properties of these algorithms are analyzed and compared. This implies that time-invariant systems (fixed dummy head) as well as time-variant systems (continuously rotated dummy head) are evaluated for convergence speed and tracking capability respectively. Furthermore, the noisy environment during measurements has to be considered. For this reason, it is necessary to simulate different additive noise levels to evaluate the algorithms’ behavior under these adverse conditions.
AB - Head-related transfer function (HRTF) is widely used for binaural sound reproduction over headphones. However, the acquisition of HRTFs using traditional measurement is usually a time-consuming task and can be only acquired at discrete directions. Recent work has shown, that the HRTF measurement can be speeded up and simplified via continuous acquisition by using an adaptive filter (identification of a time-variant system). With this method the traditional sampling and interpolation of many different positions can be avoided. There are many different adaptive filter types suitable for system identification, therefore this work analyses and compares various adaptive filter algorithms like Least-MeanSquares (LMS), Normalized-LMS (NLMS), RecursiveLeast-Squares (RLS), etc. The goal of this work is to achieve faster convergence speed and low steady-state mean squared error. Thus, the convergence and tracking properties of these algorithms are analyzed and compared. This implies that time-invariant systems (fixed dummy head) as well as time-variant systems (continuously rotated dummy head) are evaluated for convergence speed and tracking capability respectively. Furthermore, the noisy environment during measurements has to be considered. For this reason, it is necessary to simulate different additive noise levels to evaluate the algorithms’ behavior under these adverse conditions.
M3 - Paper
T2 - DAGA 2017
Y2 - 6 March 2017 through 9 March 2017
ER -