![]() |
Impulse or alien noise(IN) is a degrading cable binder. Mitigation of this noise is vital for increasing the throughput of the DSL systems. Currently, we are focussed on using the common mode signal to cancel the IN for downstream VDSL by applying signal processing techniques.
Source separation is a widely studied topic of research because of its potential applications for example, in the cocktail party problem, automatic speech recognition, music transcription etc. The difficulty of the problem increases with increasing number of sources to be separated and decreasing number of observations of the mixture. With more sources in the mixture, will focus on extraction of only one source reduce the complexity of the problem? Currently, we are trying to explore a one-at- a-time strategy for source separation in case of single observation (single channel). Both dictionary learning and neural network based methods are being used.