Neural network noise anomaly recognition system and method
First Claim
1. A computer implemented method for determining the presence or absence of a non-noise anomaly within noise by processing a received waveform, said method comprising:
- producing a plurality of samples of said received waveform;
applying said plurality of samples to one or more initial neural networks, each of said one or more initial neural networks being trained to recognize noise and only noise, said one or more initial neural networks producing one or more respective outputs related to said presence or absence of said non-noise anomaly; and
analyzing said one or more respective outputs of said one or more initial neural networks to determine if said non-noise anomaly is present in said received waveform.
1 Assignment
0 Petitions
Accused Products
Abstract
A system and method for a neural network is disclosed that is trained to recognize noise characteristics or other types of interference and to determine when an input waveform deviates from learned noise characteristics. A plurality of neural networks is preferably provided, which each receives a plurality of samples of intervals or windows of the input waveform. Each of the neural networks produces an output based on whether an anomaly is detected with respect to the noise, which the neural network is trained to detect. The plurality of outputs of the neural networks is preferably applied to a decision aid for deciding whether the input waveform contains a non-noise component. The decision aid may include a database, a computational section and a decision module. The system and method may provide a preliminary processing of the input waveform and is used to recognize the particular noise rather than a non-noise signal.
9 Citations
24 Claims
-
1. A computer implemented method for determining the presence or absence of a non-noise anomaly within noise by processing a received waveform, said method comprising:
-
producing a plurality of samples of said received waveform;
applying said plurality of samples to one or more initial neural networks, each of said one or more initial neural networks being trained to recognize noise and only noise, said one or more initial neural networks producing one or more respective outputs related to said presence or absence of said non-noise anomaly; and
analyzing said one or more respective outputs of said one or more initial neural networks to determine if said non-noise anomaly is present in said received waveform. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10)
storing said respective outputs as data sets in a database;
calculating statistical criteria for each of said data sets; and
selecting the data set having statistical criteria most determinative of said non-noise anomaly being present.
-
-
4. The method according to claim 3, wherein the statistical criteria is standard deviation.
-
5. The method of claim 1, wherein said producing a plurality of samples further comprises dividing said received waveform into one or more windows, said received waveform within each of said one or more windows being sampled and applied to a respective one of said one or more initial neural networks.
-
6. The method of claim 5, further comprising said one or more windows being incremented so as to slide relative to said received waveform with each increment.
-
7. The method of claim 6, further comprising said one or more windows being incremented until all of said received waveform is sampled.
-
8. The method of claim 1, further comprising said initial neural networks being trained to recognize Gaussian noise.
-
9. The method of claim 1, further comprising storing said respective outputs from said one or more initial neural networks in a database.
-
10. The method of claim 1, wherein said analyzing includes calculating standard deviations related to said respective outputs.
-
11. A detector for a received waveform, comprising:
-
a plurality of initial neural networks, each of said plurality of initial neural networks being programmed for recognizing noise, said plurality of initial neural networks producing a respective plurality of outputs related to the presence or absence of a non-noise anomaly; and
a decision making aid for receiving said plurality of outputs, said decision making aid being programmed to determine if a non-noise element is present or not from said plurality of outputs. - View Dependent Claims (12, 13, 14, 15, 16)
a database module storing each of said plurality of outputs as data sets;
a calculation module obtaining the standard deviation of the data sets; and
a detection module selecting the data set having the least standard deviation for comparison with a non-noise element.
-
-
15. The detector of claim 11, wherein said initial neural networks are programmed for a specific selected noise structure.
-
16. The detector of claim 11, further comprising a database within said decision making aid.
-
17. A computer implemented method for processing a received waveform containing a noise element, said method comprising:
-
training one or more initial neural networks to recognize said noise element and only said noise element;
sampling said received waveform prior to filtering out said noise element to produce one or more samples for said one or more initial neural networks;
applying said one or more samples to said one or more initial neural networks for detecting said noise element; and
producing one or more outputs responsive to said noise element. - View Dependent Claims (18, 19, 20, 21, 22, 23, 24)
storing each of said one or more output as data sets in a database;
calculating statistical criteria for each of said data sets; and
selecting the data set having statistical criteria most determinative of said anomaly being present.
-
-
22. The method according to claim 21, wherein the statistical criteria is standard deviation.
-
23. The method of claim 17, further comprising dividing said received waveform into one or more windows.
-
24. The method of claim 23, wherein said sampling further comprises sampling said windows of said received signal such that each window corresponds to each of said one or more initial neural networks.
Specification