Pulse or digital communications – Spread spectrum – Direct sequence
Reexamination Certificate
2001-07-09
2003-06-10
Chin, Stephen (Department: 2634)
Pulse or digital communications
Spread spectrum
Direct sequence
C375S267000, C375S285000, C375S343000, C375S347000, C375S349000, C455S278100, C455S296000
Reexamination Certificate
active
06577675
ABSTRACT:
TECHNICAL FIELD
The present invention relates to a new method of separating N source signals from N linear combinations of said signals using a criterion based on second order statistics. The basic idea is that every input signal (measurable signal) consists of a number of source signals. The purpose of the invention is to separate the source signals from the measurable signals using a criterion.
BACKGROUND OF THE INVENTION
In many situations, measuring and sampling of signals are performed for specific purposes as in the fields of cellular phones, data communication, medical measuring equipment, etc. At such a measuring, there are one or more primary signals and several secondary signals. The primary signals are the signals of interest for the receiver of the measurement and the secondary ones are not desired.
In a situation where a telephone is used in a car, the primary source signal can be the speaker and the secondary source signal background noise. The receiver of such a phone conversation experiences the secondary source signal as a disturbance. Therefore it is desired to reduce the secondary signal. This can be achieved by measuring the signal in the car using at least two microphones.
GSM cellular phones use a so called speech coder with which you code the voice so that a narrow bandwith digital transmission can be used for transferring the communication. In order to achieve a coding, it is required that the signals is a speaking voice, i.e. not singing, music, etc. If the primary signal is disturbed by other secondary signals such as noise, music and so on, the result of the speech coding can be very poor, even not understandable. Therefore, there is much to gain by using the suggested invention in a preprocess to the speech coding so that effects of secondary signals are reduced.
Another problem in the telecommunications area is occuring echos. Echos can appear when a four wire connection is to be converted into a two wire connection. The converting can be realized with a so called hybride. For a specific connection the echo can be insignificant, but since the communication can be connected in a variety of different ways echos can appear. The problem is solved by an echo cancellor. The present invention has the potential of achieving better performance than a conventional echo cancellor and this is without surveillance strategies.
The importance of antenna array techniques are steadily increasing and with it the problem of separating conversations increases. Normally, an antenna array consists of a number of antennas arranged in a row. These antennas can, with the help of software, be controlled so that antenna lobe is arranged in the desired direction. If two mobile units are in the same area, it can be motivated to regard the potential of the invention in order to increase the quality of the signals of the mobile units separately.
When data is transferred via an electric cable, superposition of different data sequences may occur. The effects of this can be reduced by using the present invention.
A hearing impaired person normally has problems to separate the primary signal from the secondary one with the use of a conventional hearing aid. As an example, the armature of a fluorescent tube can be a secondary signal causing a great disturbance in the hearing aid. The present invention can reduce this disturbance. A similar problem occurs when the hearing impaired person is having a conversation in a group of people. This problem is also called the cocktail party problem. The present invention can reduce the effect of this problem.
Medical measuring equipment, such as electrocardiogram (E.C.G.), can also use the invention in order to give prominence to the signal of interest at the measuring.
In an article from 1991 [1] a technique to separate source signals mixtures without memory was presented. The technique is based on the assumption that the sources are statistically independent of each other. By evaluating all possible moments of the measured signals it is theoretically possible to separate the source signals from each other. However, it is practically impossible to evaluate all such moment since in the general case there is an infinite number of such moments. The method presented in [1] aims at via first and third moments accomplish a separation. By formulating the cross moments of the measures signals a non linear system of simultaneous equation appears, which can be solved by iteration. The solution to that system is not unique since the system is non-linear. This method as such is afflicted with problems since the cross moment equations are chosen by the user, i.e. in advance it is not possible to know whether the solution represents a minimum. If the desired solution does not represent a minimum, the iterative solution will not converge to it. In an article from 1994 [2] a class of criteria for this type of problem is presented. The criteria class is called contrast functions. If the problem is formulated with assistance of a contrast function, it is known that the desired point is a maximum of the contrast function, which means that an iterative method can be formulated. This means that if the algorithm is initiated in a suitable vicinity of the desired solution the algorithm will converge to it.
Separation of dynamic source signal mixtures has been presented parallel to methods for statistical mixtures. One of the first which was presented was Adaptive Noise Cancellation (ANC) [3]. This method assumes that one of the measured signals consists of one and only one source. This assumption is not realistic in many of the problems mentioned previously. In an articel from 1985 [4] a method was presented which uses a backward demixing structure. The method for separation of measured signals is based on a connected version of ANC. In the method the square of the demixing signals s
1
and s
2
are minimized. Since the minimization takes place via two independent criteria, it can not be said that the method will solve the problem. This method is a subject of a patent application.
The problem with dynamic mixtures is also described in articles [5] and [6], of which the latter is of interest since it can potentially solve the problem via a minimization of the loglikelyhood function. This requires however, that this function can be formulated. The authors of [6] wrote 1993 an article [7] which presents a method for separating dynamically mixed signals via a decorrelation method. This method is a subject of a patent application. In the patent application it is started out from ANC and it is demonstrated how the solution seems when the requirements of ANC are fulfilled. Thereafter, a generalization is formulated of this solution. The generalization is of ad hoc character and cannot therefore be said to be correct. If the method is to separate two dynamically mixed sources with different zero weights, then the algorithm will not solve the problem. This indicates that the generalization is not correct.
Also, ANC is based on minimization of an auto correlation, and in the described generalization the desired solution can be represented by a saddle point rather than by a minimum. In such a case, convergence to the desired solution will not occur. In an article from 1994 [8], a method is described to solve the problem with dynamic channels by using higher order moment. In that article the authors claim that the method can be designed by means of a cost function. This method is a subject of a patent application. However, the method is not based on second order statistics, which results in the need of great computer computational power and more samples.
MOST CHARACTERIZING PARTS OF THE INVENTION
Characteristics of the present invention are that every measured signal is brought to a separation structure comprising cross connected linear filters and subsequent adders, or that every measured signal is brought to a separation structure comprising adders and cross connected linear fi
Broman Holger
Lindgren Ulf
Chin Stephen
Ha Dac V.
Oppedahl & Larson LLP
Telefonaktiegolaget LM Ericsson
LandOfFree
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 Signal separation, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Signal separation will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3147345