Boots – shoes – and leggings
Patent
1992-07-08
1994-05-24
Harvey, Jack B.
Boots, shoes, and leggings
G06F 1520
Patent
active
053155327
DESCRIPTION:
BRIEF SUMMARY
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to a method and a device for real-time separation of mixed signals.
2. Discussion of the Related Art
The problem posed is the blind identification of a mixture of signals. Signals are received on a certain number of sensors, equal to p in number for example. These signals come from a mixture, which is by assumption linear but whose transfer function is unknown, of a number n less than or equal to p of "source" signals which are also unknown, and which, also by assumption, come from independent and non-Gaussian (except strictly for one of them alone) sources. It therefore involves at the same time separating, that is to say identifying, these source signals and determining the transfer function effecting the mixture of signal which are received on the p abovementioned sensors: it is therefore a problem of blind deconvolution.
Taking the simple example from the field of radar of sonar, the signals received by the sensors which together constitute the reception antenna, are generally processed to form channels in directions chosen a priori, this making it possible to spatially separate "sources" constituted by echoes or noise-makers. The problem posed here is to identify there channels directly.
Thus, to fix ideas, if two sources x.sub.1 (t) and x.sub.2 (t), where t is the time variable, and two sensors providing two signals e.sub.1 (t) and e.sub.2 (t) such that: is to say to identify the transform matrix: ##EQU1## the two channels x.sub.1 (t) and x.sub.2 (t).
In this example, the source signals are taken at the same instant t: the mixture of signals is termed "instantaneous". In general, mixtures are not instantaneous but "convolutive". It is nevertheless possible to reduce a convolutive problem to an instantaneous problem by decomposing the signals into signals with pure frequencies by spectral analysis obtained by FOURIER Transformation. Denoting the frequency by f, each observed signal can then be written: ##EQU2## and each source signal can be written: ##EQU3## Finally, the blind deconvolution problem to be solved here can be stated thus:
p signals (p greater than or equal to 2) are observed, and for which it is known that they arise from n unknown source signals (n less than or equal to p) through an unknown linear, stationary transformation A(f) such that: ##EQU4## the signals e(t), x(t) and the transformation A possibly being complex data. Moreover, independent and, except strictly for one of them along, non-Gaussian sources are involved. Blind deconvolution then consists in the determination of the transfer function A and thereby of each of the source-signals.
A known solution for effecting a time separation of an instantaneous mixture A(f)=A of signals is proposed by Messrs. C. JUTTEN and J. HERAULT in several articles including the one published in the French journal "Traitement du Signal", volume 5, No. 6, 1988, pages 389 to 403. In involves a method of separation which uses a fully interconnected signal-layer array of linear neurons whose weights are supervised by an algorithm akin to that of stochastic iteration.
This known method nevertheless has some disadvantages: erroneous outputs, which would diverge if there were no natural saturation; be extremely slow; on the speed of convergence and on the solution provided.
SUMMARY OF THE INVENTION
The invention aims to remedy these disadvantages. To this effect it relates to a method of real-time separation of signals received by a predetermined number p of sensors, these signals coming from a linear mixture, but of unknown transfer function, of a number n less than or equal to p of source-signals which are also unknown and which come from independent and, except strictly for one of them, non-Gaussian sources, this method consisting in processing these received, sensed and sampled signals, in two steps, including: s(t) from these received signals e(t); and x(t) being the required source-signals, this orthogonal matrix Q, which effects a linear transformation, being
REFERENCES:
patent: 3751596 (1973-08-01), Tseng
patent: 4947480 (1990-08-01), Lewis
New Methods for Signal Separation Fety et al., 1988 (no month), pp. 226-230.
Co-Channel Interference Suppression and Signal Separation Method, Bar-Ness et al., 1988 (no month), pp. 1077-1081.
SPIE, Advanced Algorithms and Architectures for Signal Processing III, vol. 975, 1988, J. G. McWhirter et al.: "Efficient MVDR processing using a systolic array", pp. 385-392.
Traitement du Signal, vol. 5, No. 6, 1988, C. Jutten et al.: "Une solution neurominetique au probleme de separation de sources", pp. 389-403.
IEEE Transactions on Aerospace and Electronic Systems, vol. 25, No. 4, Jul. 4, 1989, IEEE, Plc., (New York, US), M. Yuen: "Algorithmic, architectural and beam pattern issues of sidelobe cancellation", pp. 459-471.
"Thomson-CSF"
Harvey Jack B.
Peeso Thomas
LandOfFree
Method and device for real-time signal separation does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method and device for real-time signal separation, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method and device for real-time signal separation will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-1978273