Bounds on error amplification in oversampled filter banks for robust transmission

This paper analyzes the problem of error amplification in frame-based coding schemes derived from oversampled filter banks. Oversampled filter banks are currently being proposed for robust transmission applications, since added redundancy can be used for error recovery in case of losses. Error amplification is determined by the lower bound of the subframe resulting after losses. Theoretical expressions are derived to characterize such lower bound for different loss statistics. Criteria are derived that can be useful for the design of oversampled filter banks in various applications. The analysis and some application examples demonstrate that filter bank performance strongly depends on how the coefficients are organized into packets