Related Experiment Videos
Time-delay estimation of reverberated speech exploiting harmonic structure
1Division of Engineering and Applied Sciences, Harvard University, Cambridge, Massachusetts 02138, USA. msb@hrl.harvard.edu
The Journal of the Acoustical Society of America
|May 21, 1999
Summary
This study introduces a novel speech signal processing method for accurate time-delay estimation. The new approach enhances generalized cross correlation (GCC) filtering for improved talker localization and beamforming in noisy environments.
Area of Science:
- Acoustics
- Signal Processing
- Speech Technology
Background:
- Accurate time-delay estimation is crucial for talker localization and microphone array beamforming.
- Traditional generalized cross correlation (GCC) methods struggle with room reverberations and background noise.
- Existing GCC filtering criteria are often specialized for either noise or multipath effects, not both.
Purpose of the Study:
- To develop a robust GCC filtering scheme for time-delay estimation.
- To address limitations of conventional GCC methods in combined noise and multipath conditions.
- To propose a signal-dependent criterion for enhanced GCC filter design.
Main Methods:
- The study proposes an alternative GCC filtering approach using signal-dependent criteria.
- The proposed method utilizes the estimated periodicity of the speech signal.
- Simulations were conducted across various room conditions to evaluate the method.
Main Results:
- The proposed signal-dependent GCC filter demonstrates improved performance.
- The method shows utility in handling combined noise and multipath distortions.
- Evaluations illustrate the effectiveness relative to conventional GCC filtering approaches.
Conclusions:
- The novel GCC filtering method offers a robust solution for time-delay estimation.
- This approach is suitable for applications requiring accurate speech signal processing in challenging acoustic environments.
- The signal-dependent criterion provides a more adaptable filtering strategy.