| 摘要 |
#CMT# #/CMT# The method involves calculating, in a recursive manner, a spectral matrix of n-by-r dimensions generating a signal space, by modifying a former spectral matrix, where n is the dimension of data vectors and r is dimension of the space. The spectral matrix is obtained by calculating the exact solution of an optimization problem constraint to a dimension subspace ranging between r+1 and n. #CMT#USE : #/CMT# Used in a microcontroller or a microprocessor for tracking a signal space of dimension less than the dimension of the vectors of data e.g. audio, video or medical or seismic imaging data, obtained from a mechanical measurement or an audio signal, over a computing and/or telephony network for automatic analysis of a music piece through a station e.g. radio or television, to determine musical rhythm and/or transcribe and to recognize music notes and/or instruments, for data e.g. audio signal, coding in data compression field, for spectral analysis of a signal e.g. audio signal (all claimed), for soundproofing of speech signal to recognize the speech and voice and to control election campaign, for separation of a sinusoidal part and a noise part of a signal, and for image restoration, where medical imaging is obtained through NMR. #CMT#ADVANTAGE : #/CMT# Permits improved high resolution spectral analysis of the temporal data at low calculation cost, and avoids resorting to a search algorithm for a solution approached in a space of n dimension. Updating of the spectral matrix can be carried out by a simple recurrence without complex matrix multiplication. Spectral matrix calculation and signal space tracking are performed rapidly. #CMT#DESCRIPTION OF DRAWINGS : #/CMT# The drawing shows a flowchart of the pseudo-code of a signal space tracking method. |
