Dual Domain Echo Cancellation for Discrete Multitone Systems

Due to the tremendous popularity of the Internet in recent years and the growing market for bandwidth intensive applications, such as video on demand, high definition television and remote collaboration, there is an enormous demand for advanced broadband access technologies. Among the available broadband solutions for fixed users, the dominant one is Digital Subscriber Line (DSL) technology where the last mile connection is provided through the twisted-pair telephone lines. In a full-duplex DSL transmission, one of the technical challenges is to overcome the deteriorating effect of the echo, which is the interference of the transmitted signal into the collocated receiver. The digital echo cancellers employ adaptive filtering techniques to estimate the echo signal and then remove it from the received signal. The use of echo cancellation in DSL systems reduces the implementation complexity, improves the spectral efficiency and increases the loop reach. For discrete multitone (DMT) modulated DSL systems, the computational complexity of the echo cancellers can be reduced by making use of the structure present in the transmitted signal. However, this generally leads to a mixed adaptation structure in which the relationship between the time-domain and frequency-domain adaptive coefficients are not fully understood.;In this thesis, we introduce a general framework for studying and designing echo cancellers for DMT-based systems. This framework establishes a unified representation of the existing echo cancellation methods and provides the capability of achieving different design trade-offs between the convergence rate and the computational complexity of the echo cancellation algorithms. This capability is also exploited to design new cancellers with improved performance and/or lower complexity. In the first part of this thesis, echo cancellation for DMT-based systems is reformulated as a constrained optimization problem, where a cost function is to be minimized over an extended linear space comprised of the adaptive filter coefficients in the time and frequency domains. Based on this formulation, a new linearly constrained echo cancellation structure is proposed. This new structure offers additional flexibility in implementation by allowing the incorporation of additional constraints to improve the performance of the system. In the second part of this thesis, the echo cancellation is examined in a more general dual domain, rather than only in the time and frequency domains. A two-step design methodology is proposed providing high level of flexibility in the design process of echo cancellers. Following this methodology, a new dual domain canceller is developed, which transforms the signals and weight vector into a dual domain. The proposed dual transform domain canceller has a faster convergence rate compared to those of existing cancellers with similar computational complexity.