Abstract: A computer system for performing an efficient inverse Discrete Fourier Transform (iDCT) is described. Single Instruction Multiple Data (SIMD) instructions are performed concurrently on a plurality of fixed-point data stored in multimedia registers. Operations on fixed-point data can be performed more quickly than corresponding floating-point operations. Throughout the processing of the iDCT, the fixed-point data are carefully shifted to retain the most significant bits and thereby preserve the accuracy of the multimedia operations. The initial step of transposing the matrix of multimedia data is accomplished while the data is collected to further improve efficiency. Moreover, the step of transposing the matrix of data is broken down into a series of smaller transpositions, depending on the amount of data on which the processor can operate at the same time.

Abstract: A noise reducing system includes an input transducer that energizes an adaptive filter. The adaptive filter energizes an output transducer that creates a noise reducing field at the input transducer.

Abstract: A system separates unknown signals which have been combined together through unknown linear filters and for which observations at multiple sensors are made. In a two channel circuit with two inputs and two sensors, the reconstructed source signals are assumed to be decorrelated such that the cross-correlation between the reconstructed source signals is near zero. The transfer functions which represent the crosstalk processes are estimated. The output signals are detected and the transfer functions are recursively solved. A reconstruction filter is used to recover the original input signals.