DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Claim Objections
Claim 12 is objected to because of the following informalities:
Claim 12 – Please change “uniform circular array” to “the uniform circular array.”
Appropriate correction is required.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Examiner’s Note: Claims 5-7, 17-19, and 20 are not rejected under this section.
Claims 1-4, 8, and 9 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more. The claim(s) recite(s) the abstract idea of a mathematical algorithm for producing an updated mathematical model for a direction finding system through a mathematical updating process (i.e., the production of an updated two-argument arctangent function model through the mathematical process of transfer learning based on sine/cosine patterns to optimize the mathematical model parameters).
This judicial exception is not integrated into a practical application because no use of the mathematical model is recited in any manner that would amount to an improvement to the underlying direction finding system.
The claim(s) does/do not include additional elements that are sufficient to amount to significantly more than the judicial exception because the recited use of the processors, recording media, and neural networks amount to the recitation of general-purpose computer elements for implementing the algorithm through use of a general-purpose computer and do not serve to amount to significantly more than the recitation of the abstract idea itself (see Alice Corp. v. CLS Bank International, 573 U.S. 208 (2014)). Accessing and outputting the model (i.e., neural network) amounts to insignificant extra-solution activity in implementing the abstract idea using a general-purpose computer.
Claims 10-16 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more. The claim(s) recite(s) the abstract idea of a mathematical algorithm determining an angle of arrival based on measurements from a direction finding sensor array.
This judicial exception is not integrated into a practical application because no improvement to the underlying direction finding system outside of the use of an improvement to the mathematics used to determine angle of arrival.
The claim(s) does/do not include additional elements that are sufficient to amount to significantly more than the judicial exception because the recited use of the processors, recording media, and neural networks amount to the recitation of general-purpose computer elements for implementing the algorithm through use of a general-purpose computer and do not serve to amount to significantly more than the recitation of the abstract idea itself (see Alice Corp. v. CLS Bank International, 573 U.S. 208 (2014)). Accessing the data needed to implement the abstract idea and outputting the algorithm results amount to the recitation of insignificant extra-solution activity in implementing the abstract idea using a general-purpose computer. The recitations regarding the arrangement of the sensor array sensors amount to mere field-of-use limitations.
Claim Rejections - 35 USC § 102
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale, or otherwise available to the public before the effective filing date of the claimed invention.
Claim(s) 1-20 is/are rejected under 35 U.S.C. 102(a)(1) as being anticipated by Friedrichs et al., Watson–Watt Direction Finding With Transfer Learning-Based Calibration, IEEE, January 2023 [hereinafter “Friedrichs”].
Examiner’s Note: See MPEP 715.01(c) regarding the filing of a “Katz Declaration”:
Where the inventor or at least one joint inventor is a co-author of a publication cited against an application, a rejection of the application under pre-AIA 35 U.S.C. 102(a) or (e) based on the publication may be overcome by filing an affidavit or declaration of the inventor or at least one joint inventor under 37 CFR 1.131(a). Alternatively, the rejection may be overcome by filing a specific affidavit or declaration of the inventor or at least one joint inventor under 37 CFR 1.132 establishing that the publication is describing work of the inventor or one or more joint inventors. An uncorroborated affidavit or declaration by a sole inventor or single joint inventor indicating that person to be the sole inventor and the other co-authors to have been merely working under their direction has been sufficient to remove the publication as a reference under pre-AIA 35 U.S.C. 102(a). In re Katz, 687 F.2d 450, 215 USPQ 14 (CCPA 1982). However, an affidavit or declaration under 37 CFR 1.132 that is only a naked assertion of inventorship by an (joint) inventor who has an interest at stake and that fails to provide any context, explanation or evidence to support that assertion is insufficient to show that the relied-upon subject matter was the work of the inventive entity. See EmeraChem Holdings, LLC v. Volkswagen Grp. of Am., Inc., 859 F.3d 1341, 1345, 123 USPQ2d 1146, 1149 (Fed. Cir. 2017) (The court found the declaration submitted by inventor Campbell more than twenty years after the invention insufficient to establish that he and Mr. Guth (deceased) were the inventors of the subject matter disclosed in a patent naming Campbell, Guth, Danziger, and Padron as inventors because "[n]othing in the declaration itself, or in addition to the declaration, provides any context, explanation, or evidence to lend credence to [Campbell’s] bare assertion." Id. at 1345, 123 USPQ2d at 1149.).
Regarding Claim 1, Friedrichs discloses a system for facilitating calibration of a direction finding system [Fig. 1], the system comprising: one or more processors; and one or more computer-readable recording media that store executable instructions that are executable by the one or more processors [Inherent to the use of the neural networks] to configure the system to:
access a baseline neural network initially configured to imitate behavior of a two-argument arctangent function [Fig. 2];
apply transfer learning to the baseline neural network to generate a calibrated neural network, wherein the transfer learning calibrates the baseline neural network to perform Watson-Watt direction finding without utilizing a lookup table for error correction [Abstract and Sections III and IV]; and
output the calibrated neural network [Conclusion – “one standardized NN system, ataNN2, can be deployed to all sensor systems”].
Regarding Claim 2, Friedrichs discloses that the baseline neural network comprises a first neural network and a second neural network [Section III].
Regarding Claim 3, Friedrichs discloses that the first neural network and the second neural network comprise separate instances of a single initially trained neural network [Section III].
Regarding Claim 4, Friedrichs discloses that the baseline neural network is configured to receive input comprising a sine pattern and a cosine pattern [Figs. 1 and 2].
Regarding Claim 5, Friedrichs discloses that the baseline neural network is configured to: process the input using the first neural network when a sign of the sine pattern is positive; and process the input using the second neural network when the sign of the sine pattern is negative [Fig. 2, Y and -Y].
Regarding Claim 6, Friedrichs discloses that processing the input using the second neural network comprises applying a sign change to the sine pattern [Fig. 2, Y and -Y].
Regarding Claim 7, Friedrichs discloses that processing the input using the second neural network comprises applying a post-processing angle transformation to output of the second neural network [Fig. 2, Out - 180].
Regarding Claim 8, Friedrichs discloses that applying transfer learning to the baseline neural network comprises retraining the first neural network and the second neural network utilizing transfer learning data [Section IV] comprising measurement data and ground truth angle of arrival data [Section V].
Regarding Claim 9, Friedrichs discloses that the transfer learning data further comprises anchor point data associated with one or more beam peaks or one or more beam crossovers [Fig. 3].
Regarding Claim 10, Friedrichs discloses a system for performing direction finding [Fig. 1], the system comprising: one or more processors; and one or more computer-readable recording media that store executable instructions that are executable by the one or more processors [Inherent to the use of the neural networks] to configure the system to:
access measurement data acquired via a direction finding sensor array [Fig. 1];
generate preprocessed data by applying one or more preprocessing operations to the measurement data [Fig. 3 and corresponding text and Section V];
utilize the preprocessed data as input to a calibrated neural network, wherein the calibrated neural network is calibrated via transfer learning to perform Watson-Watt direction finding without utilizing a lookup table for error correction; and output angle of arrival data, the angle of arrival data comprising output of the calibrated neural network [Abstract and Sections III and IV].
Regarding Claim 11, Friedrichs discloses that the direction finding sensor array comprises a uniform circular array of monopoles [Fig. 1].
Regarding Claim 12, Friedrichs discloses that that uniform circular array of monopoles comprises a four-element array [Fig. 1].
Regarding Claim 13, Friedrichs discloses that the measurement data comprises radiation pattern data, and wherein applying the one or more preprocessing operations to the measurement data comprises mapping the radiation pattern data to sine pattern data and cosine pattern data [Fig. 1 and Section V].
Regarding Claim 14, Friedrichs discloses that mapping the radiation pattern data to the sine pattern data and the cosine pattern data comprises determining maximum power of different channel pairs, changing polarity of at least one channel from each of the different channel pairs, and, after changing polarity, fitting channel pair data to a unit circle via vector norm [Section V].
Regarding Claim 15, Friedrichs discloses that the one or more preprocessing operations comprise an offset removal operation [Section V].
Regarding Claim 16, Friedrichs discloses that the calibrated neural network comprises a first calibrated neural network and a second calibrated neural network [Fig. 2 and Section IV].
Regarding Claim 17, Friedrichs discloses that the calibrated neural network is configured to: process the preprocessed data as input using the first calibrated neural network when a sign of the sine pattern data is positive; and process the preprocessed data as input using the second calibrated neural network when the sign of the sine pattern data is negative [Fig. 2, Y and -Y].
Regarding Claim 18, Friedrichs discloses that processing the preprocessed data as input using the second calibrated neural network comprises applying a sign change to the sine pattern data [Fig. 2, Y and -Y].
Regarding Claim 19, Friedrichs discloses that processing the preprocessed data as input using the second calibrated neural network comprises applying a post-process angle transformation to output of the second calibrated neural network [Fig. 2, Out - 180].
Regarding Claim 20, Friedrichs discloses a direction finding system, comprising: a direction finding sensor array [Fig. 1];
one or more processors; and one or more computer-readable recording media that store executable instructions that are executable by the one or more processors [Inherent to the use of the neural networks] to configure the direction finding system to:
access a baseline neural network initially configured to imitate behavior of a two-argument arctangent function [Fig. 2];
apply transfer learning to the baseline neural network to generate a calibrated neural network, wherein the transfer learning calibrates the baseline neural network to perform Watson-Watt direction finding without utilizing a lookup table for error correction [Abstract and Sections III and IV];
access measurement data acquired via the direction finding sensor array [Fig. 1];
generate preprocessed data by applying one or more preprocessing operations to the measurement data [Fig. 3 and corresponding text and Section V];
utilize the preprocessed data as input to the calibrated neural network; and output angle of arrival data, the angle of arrival data comprising output of the calibrated neural network [Abstract and Sections III and IV].
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claim(s) 1-4, 8-16, and 20 is/are rejected under 35 U.S.C. 103 as being unpatentable over Massarella et al. (US 20180024220 A1)[hereinafter “Massarella”] and DePoy et al. (US 20230144796 A1)[hereinafter “DePoy”].
Regarding Claim 1, Massarella discloses a system for facilitating calibration [Paragraph [0177] – “Alternatively, the incident signal angles θ.sub.k at each given frequency f.sub.k may be directly calculated using a multi-layer perceptron, or other suitable neural network, provided by the incident angle determining module 14. In this case, the inputs would be the measured values of signal power P.sub.n(f.sub.k), in which 1≦n≦N, and the network would be trained, during calibration of the system using known RF signal sources 5, using back-propagation of weight values or another suitable training method.”] of a direction finding system [See Figs. 1a, 3, 8a, 8b, and 9a and corresponding text.Paragraph [0041] – “FIG. 1a schematically illustrates an Adcock antenna array for use in Watson-Watt direction finding”], the system comprising:
one or more processors; and one or more computer-readable recording media that store executable instructions that are executable by the one or more processors to configure the system [See Fig. 3 and Paragraph [0083] – “In this example, the incident angle determining module 14 is implemented by a suitably programmed microprocessor. However, the incident angle determining module 14 may alternatively be provided by a FPGA or by a suitably programmed computer.” The program instructions are inherently stored in memory.] to:
access a baseline neural network [Paragraph [0177] – “Alternatively, the incident signal angles θ.sub.k at each given frequency f.sub.k may be directly calculated using a multi-layer perceptron, or other suitable neural network, provided by the incident angle determining module 14. In this case, the inputs would be the measured values of signal power P.sub.n(f.sub.k), in which 1≦n≦N, and the network would be trained, during calibration of the system using known RF signal sources 5, using back-propagation of weight values or another suitable training method.”] initially configured to imitate behavior of a two-argument arctangent function [Paragraph [0042] – “FIG. 1b schematically illustrates ambiguities which can arise when Watson-Watt methods are applied to signal power values”See Equation (5) which is based off of the two arguments presented by Equations (3) and (4).].
Massarella discloses the use of Watson-Watt direction finding [Paragraph [0041] – “FIG. 1a schematically illustrates an Adcock antenna array for use in Watson-Watt direction finding”] without utilizing a lookup table for error correction [Use of the neural network of Paragraph [0177] “alternatively” to the use of the look-up table mentioned elsewhere (for example, Paragraph [0095])] and the use of machine learning [Paragraph [0176] – “The hereinbefore described methods of estimating incident signal angles may be modified to incorporate machine learning methods. For example, machine learning methods may be employed to optimise, during calibration and/or during operation, corrections to compensate for bias errors which may arise in any of the hereinbefore described methods.”], but fails to disclose that the system is configured to:
apply transfer learning to the baseline neural network to generate a calibrated neural network, wherein the transfer learning calibrates the baseline neural network; and
output the calibrated neural network.
However, DePoy discloses to apply transfer learning to a baseline neural network to generate a calibrated neural network, wherein the transfer learning calibrates the baseline neural network [See the neural network of Paragraphs [0152]-[0153].Paragraph [0153] – “In some implementations, training begins by using a generic model trained to perform DoA regression in the simulation domain. The model is updated for a specific (physical) array or deployment scenario by capturing target signals from one or more test positions using the target sensor/array. Such an implementation is shown in FIG. 1 in the case where the environment 102 is simulated. These recordings are then used to generate training examples to update the model trained in simulation (e.g., transfer learning). This manual training process can be repeated until the quality of the model is suitable (e.g., meets minimum criteria for accuracy/precision, satisfies a threshold, among others). Once a minimally viable model is produced, a semi-supervised training method may be deployed in order to iteratively enhance the quality of the model without the need for additional data collection. A semi-supervised training method can include generating ground truth data based on traditional DoA methods as discussed in reference to FIG. 1.”]; and
output the calibrated neural network [Paragraph [0152] – “In addition to the utility for calibration and enhancement, the transformer neural network, such as the transformer neural network of FIGS. 5B, 6A, and 6B, serves to allow a single trained regression network to be re-used with multiple arrays. Since the regression network and the transformer network are separate and independent components of the system, switching between physical arrays only requires the inclusion of a different transformer network, instead of an entirely separate regression network.”].
It would have been obvious to use such a technique to improve the performance of the neural network and to output the resulting neural network for later use.
Regarding Claim 2, Massarella discloses that the baseline neural network comprises a first neural network and a second neural network [Paragraph [0177], each layer of the “multi-layer perceptron” considered a separate neural network as they each inherently comprise neurons].
Regarding Claim 3, Massarella discloses that the first neural network and the second neural network comprise separate instances of a single initially trained neural network [Paragraph [0177], the layers of the initially trained multi-layer perceptron prior to calibration].
Regarding Claim 4, Massarella discloses that the baseline neural network is configured to receive input [Paragraph [0177] – “Alternatively, the incident signal angles θ.sub.k at each given frequency f.sub.k may be directly calculated using a multi-layer perceptron, or other suitable neural network, provided by the incident angle determining module 14. In this case, the inputs would be the measured values of signal power P.sub.n(f.sub.k), in which 1≦n≦N, and the network would be trained, during calibration of the system using known RF signal sources 5, using back-propagation of weight values or another suitable training method.”] comprising a sine pattern [See Equation (6.1)] and a cosine pattern [See Equation (6.2)].
Regarding Claim 8, the combination would disclose that applying transfer learning [per DePoy] to the baseline neural network comprises retraining the first neural network and the second neural network [Layers of the multi-layer perceptron of Massarella] utilizing transfer learning data comprising measurement data and ground truth angle of arrival data [Paragraph [0153] – “In some implementations, training begins by using a generic model trained to perform DoA regression in the simulation domain. The model is updated for a specific (physical) array or deployment scenario by capturing target signals from one or more test positions using the target sensor/array. Such an implementation is shown in FIG. 1 in the case where the environment 102 is simulated. These recordings are then used to generate training examples to update the model trained in simulation (e.g., transfer learning). This manual training process can be repeated until the quality of the model is suitable (e.g., meets minimum criteria for accuracy/precision, satisfies a threshold, among others). Once a minimally viable model is produced, a semi-supervised training method may be deployed in order to iteratively enhance the quality of the model without the need for additional data collection. A semi-supervised training method can include generating ground truth data based on traditional DoA methods as discussed in reference to FIG. 1.”].
Regarding Claim 9, DePoy fails to disclose that the transfer learning data further comprises anchor point data associated with one or more beam peaks or one or more beam crossovers. However, Massarella discloses that the arrangement would have measurement conditions corresponding to the recited anchor point data associated with one or more beam peaks or one or more beam crossovers [See Fig. 9a]. It would have been obvious to make use of such features when performing transfer learning in order to improve the accuracy of the resulting model.
Regarding Claim 10, Massarella discloses a system for performing direction finding [See Figs. 1a, 3, 8a, 8b, and 9a and corresponding text.Paragraph [0041] – “FIG. 1a schematically illustrates an Adcock antenna array for use in Watson-Watt direction finding”], the system comprising:
one or more processors; and one or more computer-readable recording media that store executable instructions that are executable by the one or more processors [See Fig. 3 and Paragraph [0083] – “In this example, the incident angle determining module 14 is implemented by a suitably programmed microprocessor. However, the incident angle determining module 14 may alternatively be provided by a FPGA or by a suitably programmed computer.” The program instructions are inherently stored in memory.] to configure the system to:
access measurement data acquired via a direction finding sensor array [See Figs. 1a, 3, 8a, 8b, and 9a and corresponding text.Paragraph [0041] – “FIG. 1a schematically illustrates an Adcock antenna array for use in Watson-Watt direction finding”];
generate preprocessed data by applying one or more preprocessing operations to the measurement data [See Equation (2) and Paragraph [0069] – “The antennae are arranged in pairs: north-south and east-west. The signals from these antennas are processed by combining the voltage signals from the north-south pair in a passive circuit known as a 180 degree hybrid.”]; and
utilize the preprocessed data as input to a calibrated neural network [Paragraph [0177] – “Alternatively, the incident signal angles θ.sub.k at each given frequency f.sub.k may be directly calculated using a multi-layer perceptron, or other suitable neural network, provided by the incident angle determining module 14. In this case, the inputs would be the measured values of signal power P.sub.n(f.sub.k), in which 1≦n≦N, and the network would be trained, during calibration of the system using known RF signal sources 5, using back-propagation of weight values or another suitable training method.”See Equations (6.1) and (6.2), used further on in the data processing.] to perform Watson-Watt direction finding [Paragraph [0041] – “FIG. 1a schematically illustrates an Adcock antenna array for use in Watson-Watt direction finding”] without utilizing a lookup table for error correction [Use of the neural network of Paragraph [0177] “alternatively” to the use of the look-up table mentioned elsewhere (for example, Paragraph [0095])] and the use of machine learning [Paragraph [0176] – “The hereinbefore described methods of estimating incident signal angles may be modified to incorporate machine learning methods. For example, machine learning methods may be employed to optimise, during calibration and/or during operation, corrections to compensate for bias errors which may arise in any of the hereinbefore described methods.”], but fails to disclose that the calibrated neural network is calibrated via transfer learning.
However, DePoy discloses to apply transfer learning to a baseline neural network to generate a calibrated neural network, wherein the transfer learning calibrates the baseline neural network [See the neural network of Paragraphs [0152]-[0153].Paragraph [0153] – “In some implementations, training begins by using a generic model trained to perform DoA regression in the simulation domain. The model is updated for a specific (physical) array or deployment scenario by capturing target signals from one or more test positions using the target sensor/array. Such an implementation is shown in FIG. 1 in the case where the environment 102 is simulated. These recordings are then used to generate training examples to update the model trained in simulation (e.g., transfer learning). This manual training process can be repeated until the quality of the model is suitable (e.g., meets minimum criteria for accuracy/precision, satisfies a threshold, among others). Once a minimally viable model is produced, a semi-supervised training method may be deployed in order to iteratively enhance the quality of the model without the need for additional data collection. A semi-supervised training method can include generating ground truth data based on traditional DoA methods as discussed in reference to FIG. 1.”].
It would have been obvious to use such a technique to improve the performance of the neural network.
The combination would disclose to output angle of arrival data [Fig. 3, element 17. See Paragraph [0083].], the angle of arrival data comprising output of the calibrated neural network [Per updating of the neural network via DePoy].
Regarding Claim 11, Massarella discloses that the direction finding sensor array comprises a uniform circular array of monopoles [See Figs. 1a, 6a, 8a, 8b, and 9a.Paragraph [0068] – “In an Adcock antenna array 1 adapted for higher frequencies, each antenna can be replaced by a pair of vertically polarised monopole antennae, the outputs of which are summed.”Paragraph [0075] – “The antenna array 8 may include antennae 9 of a single type, or the antenna array 8 may include multiple types of antennae 9. For example, each antenna 9 may be a monopole antenna, a loop antenna, a dipole antenna, a Yagi antenna, a dish antenna or any other suitable type of antenna.”].
Regarding Claim 12, Massarella discloses that the uniform circular array of monopoles comprises a four-element array [See Figs. 1a, 6a, 8a, 8b, and 9a.Paragraph [0068] – “In an Adcock antenna array 1 adapted for higher frequencies, each antenna can be replaced by a pair of vertically polarised monopole antennae, the outputs of which are summed.”Paragraph [0075] – “The antenna array 8 may include antennae 9 of a single type, or the antenna array 8 may include multiple types of antennae 9. For example, each antenna 9 may be a monopole antenna, a loop antenna, a dipole antenna, a Yagi antenna, a dish antenna or any other suitable type of antenna.”].
Regarding Claim 13, Massarella discloses that the measurement data comprises radiation pattern data [See Figs. 1a and 9a], and wherein applying the one or more preprocessing operations to the measurement data [See Equation (2) and Paragraph [0069] – “The antennae are arranged in pairs: north-south and east-west. The signals from these antennas are processed by combining the voltage signals from the north-south pair in a passive circuit known as a 180 degree hybrid.”] comprises mapping the radiation pattern data to sine pattern data [See Equation (6.1)] and cosine pattern data [See Equation (6.2)].
Regarding Claim 14, Massarella discloses that mapping the radiation pattern data to the sine pattern data and the cosine pattern data comprises determining maximum power of different channel pairs [See Figs. (6.1) and (6.2). Paragraph [0071] – “P.sub.NS is the power of the north-south voltage signal ψ.sub.NS and P.sub.EW is the power of the east-west voltage signal ψ.sub.EW. Thus, the ratio of the powers of north-south and east-west antenna signals is proportional to tan.sup.2(θ).”],
changing polarity of at least one channel from each of the different channel pairs, and, after changing polarity, fitting channel pair data to a unit circle via vector norm [Paragraph [0069] – “The antennae are arranged in pairs: north-south and east-west. The signals from these antennas are processed by combining the voltage signals from the north-south pair in a passive circuit known as a 180 degree hybrid. The output from the 180 degree hybrid is the vector subtraction of the N and S antennas:
ψ.sub.NS(t)=ψ.sub.N(t)−ψ.sub.S(t)=2iA(t)sin(k.Math.L sin θ) (2)
in which ψ.sub.NS is the result of taking the difference of north and south voltage signals ψ.sub.N, ψ.sub.S. Equation 2 can be simplified using a small angle approximation, provided that k.Math.L.Math.sin(θ)<<1:
ψ.sub.NS=2iA(t).Math.k.Math.L sin θ (3)
With correct spacing, L, between the antenna elements, this yields an azimuth polar pattern with a figure-of eight characteristic, in which each of the two lobes of the figure of eight closely approximate a circle with maximum sensitivity along the north-south axis and cancellation nulls along the east-west axis.”Paragraph [0101] – “Each directional antenna 24 has an associated direction vector V. The direction vector V relates to a distinctive point in the respective antenna gain pattern 22 of the directional antenna 24. For example, in the case that a directional antenna 24 has the Gaussian antenna gain pattern 22 as Equation 7, the direction vector V can be aligned with the direction θ=0 corresponding to the maximum antenna gain. The first type of DF system 4a is preferably configured so that the direction vectors V of the directional antennae 24 are approximately evenly distributed across all angles. However, in an application where RF signals 2 are only anticipated/desired to be received from a reduced range of incident angles θ, the directional antennae 24 may be orientated to cover a reduced range of angles.”See Fig. 9a.].
Regarding Claim 15, Massarella discloses that the one or more preprocessing operations comprise an offset removal operation [Paragraph [0058] – “FIG. 10a schematically illustrates the effects of under compensation in generating the cardioid antenna gain patterns shown in FIGS. 9c and 9d”].
Regarding Claim 16, Massarella discloses that the calibrated neural network comprises a first calibrated neural network and a second calibrated neural network [Paragraph [0177], each layer of the “multi-layer perceptron” considered a separate neural network as they each inherently comprise neurons. As calibrated per DePoy.].
Regarding Claim 20, Massarella discloses a direction finding system, comprising: a direction finding sensor array [See Figs. 1a, 3, 8a, 8b, and 9a and corresponding text.Paragraph [0041] – “FIG. 1a schematically illustrates an Adcock antenna array for use in Watson-Watt direction finding”];
one or more processors; and one or more computer-readable recording media that store executable instructions that are executable by the one or more processors [See Fig. 3 and Paragraph [0083] – “In this example, the incident angle determining module 14 is implemented by a suitably programmed microprocessor. However, the incident angle determining module 14 may alternatively be provided by a FPGA or by a suitably programmed computer.” The program instructions are inherently stored in memory.] to configure the direction finding system to:
access a baseline neural network [Paragraph [0177] – “Alternatively, the incident signal angles θ.sub.k at each given frequency f.sub.k may be directly calculated using a multi-layer perceptron, or other suitable neural network, provided by the incident angle determining module 14. In this case, the inputs would be the measured values of signal power P.sub.n(f.sub.k), in which 1≦n≦N, and the network would be trained, during calibration of the system using known RF signal sources 5, using back-propagation of weight values or another suitable training method.”] initially configured to imitate behavior of a two-argument arctangent function [Paragraph [0042] – “FIG. 1b schematically illustrates ambiguities which can arise when Watson-Watt methods are applied to signal power values”See Equation (5) which is based off of the two arguments presented by Equations (3) and (4).].
Massarella discloses the use of Watson-Watt direction finding [Paragraph [0041] – “FIG. 1a schematically illustrates an Adcock antenna array for use in Watson-Watt direction finding”] without utilizing a lookup table for error correction [Use of the neural network of Paragraph [0177] “alternatively” to the use of the look-up table mentioned elsewhere (for example, Paragraph [0095])] and the use of machine learning [Paragraph [0176] – “The hereinbefore described methods of estimating incident signal angles may be modified to incorporate machine learning methods. For example, machine learning methods may be employed to optimise, during calibration and/or during operation, corrections to compensate for bias errors which may arise in any of the hereinbefore described methods.”], but fails to disclose that the system is configured to:
apply transfer learning to the baseline neural network to generate a calibrated neural network, wherein the transfer learning calibrates the baseline neural network.
However, DePoy discloses to apply transfer learning to a baseline neural network to generate a calibrated neural network, wherein the transfer learning calibrates the baseline neural network [See the neural network of Paragraphs [0152]-[0153].Paragraph [0153] – “In some implementations, training begins by using a generic model trained to perform DoA regression in the simulation domain. The model is updated for a specific (physical) array or deployment scenario by capturing target signals from one or more test positions using the target sensor/array. Such an implementation is shown in FIG. 1 in the case where the environment 102 is simulated. These recordings are then used to generate training examples to update the model trained in simulation (e.g., transfer learning). This manual training process can be repeated until the quality of the model is suitable (e.g., meets minimum criteria for accuracy/precision, satisfies a threshold, among others). Once a minimally viable model is produced, a semi-supervised training method may be deployed in order to iteratively enhance the quality of the model without the need for additional data collection. A semi-supervised training method can include generating ground truth data based on traditional DoA methods as discussed in reference to FIG. 1.”].
Massarella, as modified, would disclose that the system is configured to:
access measurement data acquired via the direction finding sensor array [See Figs. 1a, 3, 8a, 8b, and 9a and corresponding text.Paragraph [0041] – “FIG. 1a schematically illustrates an Adcock antenna array for use in Watson-Watt direction finding”];
generate preprocessed data by applying one or more preprocessing operations to the measurement data [See Equation (2) and Paragraph [0069] – “The antennae are arranged in pairs: north-south and east-west. The signals from these antennas are processed by combining the voltage signals from the north-south pair in a passive circuit known as a 180 degree hybrid.”];
utilize the preprocessed data as input to the calibrated [Per the calibration of DePoy] neural network [Paragraph [0177] – “Alternatively, the incident signal angles θ.sub.k at each given frequency f.sub.k may be directly calculated using a multi-layer perceptron, or other suitable neural network, provided by the incident angle determining module 14. In this case, the inputs would be the measured values of signal power P.sub.n(f.sub.k), in which 1≦n≦N, and the network would be trained, during calibration of the system using known RF signal sources 5, using back-propagation of weight values or another suitable training method.”See Equations (6.1) and (6.2), used further on in the data processing.]; and
output angle of arrival data [Fig. 3, element 17. See Paragraph [0083].], the angle of arrival data comprising output of the calibrated neural network [Per updating of the neural network via DePoy].
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure:
US 20190346532 A1 – METHOD FOR DIRECTION FINDING OF AT LEAST ONE STATIONARY AND/OR MOVING TRANSMITTER AS WELL AS SYSTEM FOR DIRECTION FINDING
US 20160018509 A1 – ELECTRICALLY SMALL, RANGE AND ANGLE-OF-ARRIVAL RF SENSOR AND ESTIMATION SYSTEM
Friedrichs et al., A Compact Machine Learning Architecture for Wideband Amplitude-Only Direction Finding, IEEE, July 2022
Oestreich et al., Miniaturized Watson-Watt Direction Finder: An Advancement in Vehicle Safety, IEEE, 2012
Scorrano et al., Estimation of the Direction-of-Arrival of Incoming EM Wavefronts through a Neural Network Approach, PIERS, 2017
Scorrano et al., Compact Direction Finding Array for Tactical Aircraft Radios Through Artificial Neural Networks Estimator, IEEE, 2018
Any inquiry concerning this communication or earlier communications from the examiner should be directed to KYLE ROBERT QUIGLEY whose telephone number is (313)446-4879. The examiner can normally be reached 9AM-5PM EST.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Arleen Vazquez can be reached at (571) 272-2619. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/KYLE R QUIGLEY/Primary Examiner, Art Unit 2857
Prosecution Insights