Stochastic vector based network echo cancellation

A method for echo noise cancellation in a network, comprising the steps of:

• (A) generating a first sample by filtering an input vector through a filter based on both (i) a filter vector and (ii) an intermediate vector, each of a plurality of components in said intermediate vector having a respective random value;
• (B) generating a second sample as a difference between a third sample and said first sample, said third sample being received from said network as an echo;
• and
• (C) updating a subset of a plurality of taps of said filter based on a corresponding one of said components of said intermediate vector that has a particular value, wherein a number of elements of said subset is greater than a predetermined minimum number.

An apparatus comprising:

• means for generating a first sample by filtering an input vector through a filter based on both (i) a filter vector and (ii) an intermediate vector, each of a plurality of components in said intermediate vector having a respective random value;
• means for generating a second sample as a difference between a third sample and said first sample, said third sample being received from a network as an echo;
• and
• means for updating a subset of a plurality of taps of said filter based on a corresponding one of said components of said intermediate vector that has a particular value, wherein a number of elements of said subset is greater than a predetermined minimum number.