Conexant Systems, Inc.

Nir Shapira

Nir Tal

Ron Cohen

The present invention discloses a novel scheme of performing frequency domain filtering that does not require a Hilbert transform and which is faster than equivalent FIR filtering schemes. Such a scheme has applications in implementing filters in voice band modems on OSP platforms and in numerous other applications as well. The scheme of the present invention is shown to be superior in term of CPU resource consumption and memory storage space when compared with an equivalent implementation using prior art time domain convolution techniques. The scheme of the present invention permits the delay buffer in the far end echo canceler to store the signal in real format rather than complex format, thus reducing memory requirements by half. The present invention implements a real to analytic fast convolution in the frequency domain which uses zero CPU resources.

Method of frequency domain filtering employing a real to analytic transform

A novel system by which the utilization of a central processing unit (CPU) in performing filtering operations can be reduced by shortening the filter's length thus degrading the performance of the system down to a predetermined level or threshold. The present invention is applicable to such systems that incorporate filters whereby shortening their length decreases the performance of the system and to such systems where a reliable quality criteria exists that can be measured during run time. A method iteratively minimizes the filter's length so that the quality criteria does not fall below a predetemined threshold level. In addition, a signal to noise ratio (SN) criteria is suggested for estimating the quality of the reception of communication signals. An implementation is suggested for the method in the particular case of an echo cancellation adaptive filter. In addition, a method for determining an immediate approximation of the echo canceler filter's length as opposed to finding it iteratively.

Growing community of inventors

Ramat Hasharon, Israel

Nir Shapira