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 .
Status of Claims
Claims 1-20 are pending of which claims 1, 19 and 20 are in independent form.
Claims 1-20 are rejected under 35 U.S.C. 101.
Claims 1-20 are rejected under 35 U.S.C. 103.
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.
Claims 1-20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to a judicial exception (i.e., a law of nature, a natural phenomenon, or an abstract idea) without significantly more.
The claim(s) recite(s) network access using RF signal analysis through ML models.
With respect to step 1 of the patent subject matter eligibility analysis, the claims are directed to a process, machine, manufacture, or composition of matter.
Independent claim 1 is directed to a method, which is a process.
Independent claim 19 is directed to a system, comprising one or more processors, and one or more non-transitory, computer-readable storage media, which is directed to one of the four statutory subject matters.
Independent claim 20 directed to One or more non-transitory, computer-readable storage media, which is directed to one of the four statutory subject matters.
Independent All other claims depend on claims 1, 19, and 20. As such, claims 1-20 are directed to a statutory category.
Regarding claims 1, 19 and 20:
With respect to step 2A, prong one (Judicial Exception), the claims recite an abstract idea, law of nature, or natural phenomenon. Specifically, the following limitations recite mathematical concepts and/or mental processes and/or certain methods of organizing human activity.
The claim recites the following limitations directed to an abstract idea:
Obtaining samples of RF uplink data signals,
Providing the samples as input to a machine learning model,
Obtaining recovered data based on the ML model output,
Sending the recovered data to a destination device.
These steps collectively amount to data acquisition, analysis, and transmission, implemented using a machine learning model. Such concepts fall within recognized abstract idea categories:
Mathematical concepts/algorithms (ML models operating on signal data);
Mental process/data analysis (analyzing samples to recover data);
Information processing and transmission.
Using ML model, without reciting a specific improvement to computer or network technology, is treated as an algorithm applied to data, which is abstract. See MPEP 2106.04(a) (See Electric Power Group; Digitech, SAP v. InvestPic).
With respect to step 2A, Prong Two (Particular Application), the claims do not recite additional elements that integrate the judicial exception into a practical application. The following limitations are considered “additional elements” and explanation will be given as to why these “additional elements” do not integrate the judicial exception into a practical application.
The claims do not integrate the abstract idea into a practical application:
The claim does not recite a specific ML architecture, training technique, or signal processing improvements.
The ML models are recited at high level of generality.
The radio unit, computer system, and destination device are generic network components performing their expected functions.
The claim merely uses ML as a tool to process RF data, rather than improving RF signal reception, demodulation, or network operations in a technically specific way.
There are no: improved RF hardware performance; reducing noise, latency, or bandwidth usage via a specific technique; non-conventional signal recovery mechanism tied to physical layer constraints.
There is no recitation of, a new data structure that changes computer operation, improved network functioning, an unconventional indexing technique, a specific hardware solution. Instead, the claims recite conventional and generic computer functions performed in a routine manner, which does not amount to a practical application.
With respect to Step 2B. The claims do not include additional elements that are sufficient to amount to significantly more than the judicial exception. The recited components are merely generic computer/database elements performing their routine, well-understood, and conventional functions. See Alive, MPEP 2016.05(d).
The steps mentioned: generic computer system; generic rf SIGNAL SAMPLING; Broadly claimed ML model; conventional data transmission. These elements merely amount to applying an abstract idea using conventional computing and networking technology. Courts have consistently helped such high-level information management operations are conventional.
Considering claims as a whole, the ordered combination of elements also reflects nothing more than the typical workflow of distributed systems, and therefore DOES NOT add “significantly more” than the abstract idea.
Such generic, high‐level, and nominal involvement of a computer or computer‐based elements for carrying out the invention merely serves to tie the abstract idea to a particular technological environment, which is not enough to render the claims patent‐eligible, as noted at pg.74624 of Federal Register/Vol. 79, No. 241, citing Alice, which in turn cites Mayo. Further, See, e.g., Alice Corp. Pty. Ltd. v. CLS Bank Int'l, 134 S. Ct. 2347, 2359‐60, 110 USPQ2d 1976, 1984 (2014). See also OIP Techs. v. Amazon.com, 788 F.3d 1359, 1364, 115 USPQ2d 1090, 1093‐94 (Fed. Cir. 2015) ("Just as Diehr could not save the claims in Alice, which were directed to 'implement[ing] the abstract idea of intermediated settlement on a generic computer', it cannot save O/P's claims directed to implementing the abstract idea of price optimization on a generic computer.") (citations omitted). See also, Affinity Labs of Texas LLC v. DirecTV LLC, 838 F.3d 1253, 1257‐1258 (Fed. Cir. 2016) (mere recitation of a GUI does not make a claimpatent‐eligible); Intellectual Ventures I LLC v. Capital One Bank, 792 F.3d 1363, 1370 (Fed. Cir. 2015) ("the interactive interface limitation is a generic computer element".).
The additional elements are broadly applied to the abstract idea at a high level of generality ("similar to how the recitation of the computer in the claims in Alice amounted to mere instructions to apply the abstract idea of intermediated settlement on a generic computer,") as explained in MPEP § 2106.05(f)) and they operate in a well‐understood, routine, and conventional manner.
MPEP § 2106.0S(d)(II) sets forth the following:
The courts have recognized the following computer functions as well-understood, routine, and conventional functions when they are claimed in a merely generic manner (e.g., at a high level of generality) or as insignificant extra-solution activity.
• Receiving or transmitting data over a network, e.g., using the Internet to gather data, Symantec ... ; TLI Communications LLC v. AV Auto. LLC ... ; OIP Techs., Inc., v. Amazon.com, Inc ... ; buySAFE, Inc. v. Google, Inc ... ;
• Performing repetitive calculations, Flook ... ; Bancorp Services v. Sun Life ... ;
• Electronic recordkeeping, Alice Corp ... ; Ultramercial ... ;
• Storing and retrieving information in memory, Versata Dev. Group, Inc. v. SAP Am., Inc ... ;
• Electronically scanning or extracting data from a physical document, Content Extraction and Transmission, LLC v. Wells Fargo Bank ... ; and
• A web browser's back and forward button functionality, Internet Patent
• Corp. v. Active Network, Inc. ...
. . . Courts have held computer-implemented processes not to be significantly more than an abstract idea (and thus ineligible) where the claim as a whole amounts to nothing more than generic computer functions merely used to implement an abstract idea, such as an idea that could be done by a human analog (i.e., by hand or by merely thinking).
In addition, when taken as an ordered combination, the ordered combination adds nothing that is not already present as when the elements are taken individually. There is no indication that the combination of elements integrate the abstract idea into a practical application. Their collective functions merely provide conventional computer implementation. Therefore, when viewed as a whole, these additional claim elements do not provide meaningful limitations to transform the abstract idea into a practical application of the abstract idea or that the ordered combination amounts to significantly more than the abstract idea itself.
The dependent claims have been fully considered as well, however, similar to the findings for claims above, these claims are similarly directed to the “Mental Processes” grouping of abstract ideas set forth in the 2019 PEG, without integrating it into a practical application and with, at most, a general purpose computer that serves to tie the idea to a particular technological environment, which does not add significantly more to the claims. The ordered combination of elements in the dependent claims (including the limitations inherited from the parent claim(s)) add nothing that is not already present as when the elements are taken individually. There is no indication that the combination of elements improves the functioning of a computer or improves any other technology. Their collective functions merely provide conventional computer implementation. Accordingly, the subject matter encompassed by the dependent claims fails to amount to significantly more than the abstract idea.
Looking at the claim as a whole does not change this conclusion and the claim is ineligible.
Regarding claim 2,
The claim recites:
Sending recovered RF uplink data to one or more external computer systems.
This is merely a destination of data transmission (routine post solution activity). Data routing or forwarding does not alter the abstract nature of the underlaying data processing.
This does not change the nature of the abstract idea. It does not add a technical improvement to an abstract idea, such as improving computer functionality, data structure, or processing architecture.
There is no practical application, and no inventive step, the claims are still considered abstract.
Regarding claim 3,
The claim recites:
Receiving downlink data and controlling RF downlink transmission encoding the downlink data.
This is merely bidirectional data flow (uplink recovery and downlink transmission) using generic RF communication steps. These are conventional radio network operations and do not recite a specific improvement to RF signals or protocol operation.
This does not change the nature of the abstract idea. It does not add a technical improvement to an abstract idea, such as improving computer functionality, data structure, or processing architecture.
There is no practical application, and no inventive step, the claims are still considered abstract.
Regarding claim 4,
The claim recites:
First ML model for channel estimation; second ML model for symbol demapping.
This is merely reciting functional ML roles at a high level. Channel estimation and symbol demapping are long-standing signal processing tasks, and replacing known algorithms with ML models does not itself constitute a technical improvement absent structural or performance-based constraints.
This does not change the nature of the abstract idea. It does not add a technical improvement to an abstract idea, such as improving computer functionality, data structure, or processing architecture.
There is no practical application, and no inventive step, the claims are still considered abstract.
Regarding claim 5,
The claim recites:
Channel estimates as ML output;
Transforming samples based on channel estimates;
Feeding transformed samples into second ML models.
This is merely reciting data transformations and sequencing; these are still algorithmic processing steps operating on signal data. There are no specific non-conventional transformation, hardware interaction, or physical layer constraint is recited.
This does not change the nature of the abstract idea. It does not add a technical improvement to an abstract idea, such as improving computer functionality, data structure, or processing architecture.
There is no practical application, and no inventive step, the claims are still considered abstract.
Regarding claim 6,
The claim recites:
Channel estimates comprise a channel tensor.
Specifying a data structure format (tensor) does not change the abstract nature of the computation. Tensors are mathematical constructs commonly used in ML.
This does not change the nature of the abstract idea. It does not add a technical improvement to an abstract idea, such as improving computer functionality, data structure, or processing architecture.
There is no practical application, and no inventive step, the claims are still considered abstract.
Regarding claim 7,
The claim recites:
Recovered data comprises inferred bits.
Inferred bits merely describes the semantic meaning of output data, not a technical mechanism. This remains an abstract result of data interpretation.
This does not change the nature of the abstract idea. It does not add a technical improvement to an abstract idea, such as improving computer functionality, data structure, or processing architecture.
There is no practical application, and no inventive step, the claims are still considered abstract.
Regarding claim 8,
The claim recites:
Joint training of ML models using RF resource grids and ground-truth labels.
Training ML models using labeled data and loss function is well-understood, conventional ML practice. No specific improvement to train efficiency, convergence, or RF system operation is recited.
This does not change the nature of the abstract idea. It does not add a technical improvement to an abstract idea, such as improving computer functionality, data structure, or processing architecture.
There is no practical application, and no inventive step, the claims are still considered abstract.
Regarding claim 9,
The claim recites:
ML model configured to receive inputs of varying sizes.
Handling variable input sizes in a common ML design feature and does not improve computer functionality or RF signaling.
This does not change the nature of the abstract idea. It does not add a technical improvement to an abstract idea, such as improving computer functionality, data structure, or processing architecture.
There is no practical application, and no inventive step, the claims are still considered abstract.
Regarding claim 10,
The claim recites:
ML model comprises a fully convolutional neural network.
Identifying a CNN architecture without technical constraints or hardware interaction merely limits the abstract idea to a known ML model type.
This does not change the nature of the abstract idea. It does not add a technical improvement to an abstract idea, such as improving computer functionality, data structure, or processing architecture.
There is no practical application, and no inventive step, the claims are still considered abstract.
Regarding claim 11,
The claim recites:
RF uplink data samples provided in OFDM resource grid form.
OFDM resource grids are standard representations in wireless communications. The claim does not recite an improvement to OFDM signaling, decoding, or spectral efficiency.
This does not change the nature of the abstract idea. It does not add a technical improvement to an abstract idea, such as improving computer functionality, data structure, or processing architecture.
There is no practical application, and no inventive step, the claims are still considered abstract.
Regarding claim 12,
The claim recites:
Samples corresponds to a subset of an uplink resource grid for an RF burst.
Selecting a subset of data corresponding to a burst in routine signal segmentation and does not integrate the abstract idea into a practical application.
This does not change the nature of the abstract idea. It does not add a technical improvement to an abstract idea, such as improving computer functionality, data structure, or processing architecture.
There is no practical application, and no inventive step, the claims are still considered abstract.
Regarding claim 13,
The claim recites:
Executing the ML model in an L1 layer of a distributed unit (DU).
Merely specifying where software executes within network architecture does not change the nature of the computation or provide a technical improvement.
This does not change the nature of the abstract idea. It does not add a technical improvement to an abstract idea, such as improving computer functionality, data structure, or processing architecture.
There is no practical application, and no inventive step, the claims are still considered abstract.
Regarding claim 14,
The claim recites:
Training ML model using loss function based on pilot values and ground truth.
Pilot-based training and loss function are conventional RF and ML techniques. No unconventional training method or system level improvement is recited.
This does not change the nature of the abstract idea. It does not add a technical improvement to an abstract idea, such as improving computer functionality, data structure, or processing architecture.
There is no practical application, and no inventive step, the claims are still considered abstract.
Regarding claim 15,
The claim recites:
Training ML model using data values and ground truth values.
This is a generic supervised learning formulation, which is a mathematical optimization process. No unconventional training method or system level improvement is recited.
This does not change the nature of the abstract idea. It does not add a technical improvement to an abstract idea, such as improving computer functionality, data structure, or processing architecture.
There is no practical application, and no inventive step, the claims are still considered abstract.
Regarding claim 16,
The claim recites:
Simulating channel effects on the ground truth values.
Channel simulation is a generic mathematical modeling step, commonly used in RF system design and ML training. No unconventional training method or system level improvement is recited.
This does not change the nature of the abstract idea. It does not add a technical improvement to an abstract idea, such as improving computer functionality, data structure, or processing architecture.
There is no practical application, and no inventive step, the claims are still considered abstract.
Regarding claim 17,
The claim recites:
Patch-based processing of the RF samples.
Patch-based processing is a known ML data-partitioning technique and does not provide a technical improvement to RF reception or computing hardware.
This does not change the nature of the abstract idea. It does not add a technical improvement to an abstract idea, such as improving computer functionality, data structure, or processing architecture.
There is no practical application, and no inventive step, the claims are still considered abstract.
Regarding claim 18,
The claim recites:
Architecture includes one of: a non-batch norm, or a Smooth ReLU activation function.
Activation functions and normalization layers are purely mathematical model parameters and do not improved computer or network operation.
This does not change the nature of the abstract idea. It does not add a technical improvement to an abstract idea, such as improving computer functionality, data structure, or processing architecture.
There is no practical application, and no inventive step, the claims are still considered abstract.
Claim Rejections - 35 USC § 103
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) 1-20 are rejected under 35 U.S.C. 103 as being unpatentable over WEST; Nathan et al. (US 20200143279 A1) [West] in view of O`Shea; Timothy James et al. (US 20200343985 A1) [O’Shea].
Regarding claims 1, 19 and 20, West discloses, obtaining, by a computer system, samples of radio-frequency (RF) uplink data signals received wirelessly at a radio unit of a radio access network (In some implementations, a sample of electromagnetic energy processed by one or more radio frequency (RF) communication receivers is received from the one or more receivers. The sample of electromagnetic energy is examined to detect one or more RF signals present in the sample. In response to detecting one or more RF signals present in the sample, the one or more RF signals are extracted from the sample, and time and frequency bounds are estimated for each of the one or more RF signals. For each of the one or more RF signals, at least one of a type of a signal present, or a likelihood of signal being present, in the sample is classified [Abstract]. Also see ¶ [0006], [0009], [0021]);
providing, by the computer system, the samples of the RF uplink data signals as input to at least one machine learning model (In one general aspect, a method includes: receiving a training data set including a plurality of time-series radio frequency (RF) signal samples; generating a training vector from the plurality of time-series signal samples; generating one or more new training vectors by modifying the plurality of time-series signal samples; and using the training vector and the one or more new training vectors to train a machine learning model for at least one of RF band segmentation, RF signal detection, or RF signal labelling ¶ [0021], [0023], [0139]);
in response to providing the samples of the RF uplink data signals as input to the at least one machine learning model, obtaining, by the computer system, based on an output of the at least one machine learning model (In some implementations, generating the one or more bounding boxes includes generating one or more bounding boxes based on output of a neural network or a clustering algorithm that receives the signal class probabilities as input ¶ [0016]. Systems and techniques are disclosed herein detect, separate, and classify signals by operating on a time-frequency basis. In some implementations, a system includes a trained machine learning (“ML”) model such as a neural network, and uses the ML model in detecting, separating, and/or classifying the signals. The time-frequency basis may be a spectrogram computing technique using a discrete Fourier transform or other matrix-transformation methods. The ML model may be trained to output probability for each time-frequency bin that a particular type of signal is present in that bin, resulting in a grid of bins. The grid of bins can then be processed into bounding boxes by locating regions of similar signal probabilities ¶ [0039]).
However, West does not explicitly facilitate recovered data of the RF uplink data signals; and sending, by the computer system, the recovered data of the RF uplink signals to a destination device.
O’Shea discloses, recovered data of the RF uplink data signals (FIG. 2B is a diagram showing an example of improved error vector magnitude (EVM) upon using a machine-learning network for processing digital communications. The figure presents a performance comparison between a conventional approach to estimation and equalizing involving MMSE algorithms (plot 230) and a machine-learning network approach to the same task (plot 240). Plot 230 illustrates the recovered data symbol tiles produced by using MMSE, which involves multiplication of the estimated channel inverses from the network with the received symbol value tiles. Plot 240 illustrates recovered data symbol tiles produced by a machine-learning network approach as discussed with respect to FIGS. 1 and 2A ¶ [0086], [0136]); and
sending, by the computer system, the recovered data of the RF uplink signals to a destination device (FIG. 5 is an example of signal processing stages that typically take place within the RU, DU, or CU within an O-Ran or vRAN system for 5G-NR, 4G, or 5G+ deployment, where several options for splits between the RU and DU are possible given by options 8, 7*, and 6. 7 bears an asterisk in FIG. 5 as the option 7 split, as shown in item 520, generally includes 3 different common splits labeled 7-3, 7-2, and 7-1. Each split has different bandwidth and latency requirements between elements and varies per frequency-bandwidth, number of antenna elements and other parameters. Splits 7-2, 7-3 and similar are adopted by deployed virtual RAN (vRan) systems. The machine-learning approach can be targeted to the requirements and efficiency of the resulting DU within such a split but can similarly be used for a variety of split configurations. The split in the PHY layer typically concerns allotting processing stages to either a RU or DU within a system. FIG. 5 shows how a machine-learning network can fit into the communications system 500. In some cases, the machine learning network can reside instead within the RU, where processing stages before or after the FFT, inverse FFT (IFFT), or other processing stages including synchronization may be determined by these machine learning networks and updated based on signal quality metrics generated locally or passed as feedback from the DU or CU ¶ [0112]. Also see ¶ [0129]-[0131] and [0135]).
It would have been obvious to one ordinary skilled in the art before the effective filing date of the claimed invention to combine the teachings of the cited references because O’Shea’s system would have allowed West to facilitate recovered data of the RF uplink data signals; and sending, by the computer system, the recovered data of the RF uplink signals to a destination device. The motivation to combine is apparent in the West’s reference, because there is a need to improve communications systems that use machine learning and includes processing of communications signals using a machine-learning network.
Regarding claim 2, the combination of West and O’Shea discloses, wherein sending the recovered data of the RF uplink signals to the destination device comprises sending the recovered data of the RF uplink signals to one or more computer systems external to the radio access network (West: The processing unit 130 may save this classification and/or value regression as metadata and may store it on the data storage 120. For example, the processing unit 130 may transfer the matched classification and/or value regression data to one or more external devices through a network port as metadata 122. The processing unit 130 may save the matched classification and/or value regression data locally on a file disk as metadata 124 ¶ [0062]).
Regarding claim 3, the combination of West and O’Shea discloses, receiving, at the computer system, downlink data for a user device in response to sending the recovered data of the RF uplink signals to the destination device; and controlling, by the computer system, transmission of an RF data downlink signal from the radio unit to the user device, the RF data downlink signal encoding the downlink data (O’Shea: In some implementations, the input data 110 is an OFDM or CP-OFDM signal (e.g., in the 3GPP 5G-NR uplink (UL) or downlink (DL) PHY), as shown graphically as an input plot 111 that is a time-frequency spectrum grid of pilot and data subcarriers and time slots within an OFDM signal block ¶ [0047]. In some cases, a 64-quadrature amplitude modulation (QAM) 5G-NR system, or other fixed or learned modulation schemes, transmitted over the air in the normal physical downlink shared channel (PDSCH) or physical uplink shared channel (PUSCH) form, can be received using the conventional MMSE, LMMSE, or similar case, in order to achieve a certain bit error rate (BER) or error vector magnitude (EVM) ¶ [0105]. Also see ¶ [0121], [0127]).
Regarding claim 4, the combination of West and O’Shea discloses, wherein the at least one machine learning model comprises: a first machine learning model configured to perform channel estimation based on the samples of the RF uplink data signals (O’Shea: In some implementations, the set of ground truths are values of equalized data symbols or channel estimates determined from one or more of a process of generating the pilot and data information, a decision feedback process, pilot subcarriers, or an out-of-band communication ¶ [0012], [0034], [0086]); and
a second machine learning model configured to perform symbol-demapping on estimated symbols of the RF uplink data signals, wherein the estimated symbols are based on the channel estimation by the first machine learning model (O’Shea: The device 208 uses the output prediction from the machine-learning network 212 for detecting symbols, using symbol detection component 216, where the detected symbols are estimates of the symbols transmitted from the device 201, using the prediction output by the machine-learning network 212. The detected symbols are used in performance analysis 218 ¶ [0067], [0069], [0082]. The process can begin, as shown in the example of FIG. 5, with an RF signal being received. Pre-processing stages such as cyclic prefix removal, fast Fourier transform (FFT), port reduction and resource element de-mapping, can all be used to generate a form of an unequalized resource grid. The unequalized resource grid from pre-processing steps can be sent to a machine-learning network 522 of the system 500. In this case, the machine-learning network 522 performs channel estimation, diversity or other antenna combining, and equalization before sending the resulting equalized symbols to other elements in the process flow including inverse discrete Fourier transform (iDFT) ¶ [0115]).
Regarding claim 5, the combination of West and O’Shea discloses, wherein providing the samples of the RF uplink data signals as input to the at least one machine learning model comprises: providing the samples of the RF uplink data signals as input to the first machine learning model (West: the system includes a machine learning model that is trained using samples that are received from one or more radio receivers or simulated throughout a signal environment simulator and that are saved to a dataset. The dataset may be labelled as part of a dataset for training machine-learned models ¶ [0005], [0007], [0013], [0021], [0023]);
obtaining, as an output of the first machine learning model, channel estimates characterizing channel effects on the RF uplink data signals (O’Shea: The machine-learning network 120 learns to accomplish these tasks jointly and, in doing so, learns to compensate for channel effects and to interpolate the channel response estimate properly across a sparse grid, in some cases leveraging both data aided (e.g., reference) and non-data aided (e.g., non-reference) resource elements ¶ [0053], [0153]);
transforming the samples of the RF uplink data signals based on the channel estimates (West: Implementations may include one or more of the following features. For example, in some implementations, estimating time and frequency bounds for each of the one or more RF signals includes: transforming the sample of electromagnetic energy from a time-series representation to a time-frequency representation ¶ [0010], [0053]);
providing the transformed samples of the RF uplink data signals as input to the second machine learning model (O’Shea: These incremental changes, such as transmitter adaptation of shaping, modulation, or pre-coding, may occur as feedback mechanisms within a system including a machine-learning network (e.g., where the channel estimation can be exploited or transformed to improve or implement these functions) ¶ [0042]. The machine-learning network 212 includes several fully connected layers (FC), with a layer performing matrix multiplications of an input vector with a weight vector followed by summation to produce an output vector. In some implementations, the machine-learning network 212 includes non-linearity, such as a rectified linear unit (ReLU), sigmoid, parametric rectified linear unit (PReLU), MISH neural activation function, SWISH activation function, or other non-linearity. In some cases, the machine-learning network 212 leverages convolutional layers, skip connections, transformer layers, recurrent layers, residual layers, upsampling or downsampling layers, or a number of other techniques that serve to improve the performance of the machine-learning network 212, for example by achieving an improved performance architecture. In some implementations, the machine-learning network 212 is a convolutional neural network ¶ [0064]); and
obtaining, as an output of the second machine learning model, data indicative of the recovered data (West: In some implementations, generating the one or more bounding boxes includes generating one or more bounding boxes based on output of a neural network or a clustering algorithm that receives the signal class probabilities as input ¶ [0016]. Systems and techniques are disclosed herein detect, separate, and classify signals by operating on a time-frequency basis. In some implementations, a system includes a trained machine learning (“ML”) model such as a neural network, and uses the ML model in detecting, separating, and/or classifying the signals. The time-frequency basis may be a spectrogram computing technique using a discrete Fourier transform or other matrix-transformation methods. The ML model may be trained to output probability for each time-frequency bin that a particular type of signal is present in that bin, resulting in a grid of bins. The grid of bins can then be processed into bounding boxes by locating regions of similar signal probabilities ¶ [0039]).
Regarding claim 6, the combination of West and O’Shea discloses, wherein the channel estimates comprise a channel tensor (O’Shea: In some implementations, processing may run on a DU. For example, the DU can be a generic server platform, for instance an Intel server platform, which may employ one or more accelerators for instance it may use field-programmable gate array (FPGA) offload for the error correction decoding. The DU may use a graphic processing unit (GPU) or a Tensor or Vector processor or other systolic array to perform the neural network operations in order to reduce power consumption and to improve throughput allowing a system, such as the system 500, to perform decoding of more sectors, more radio units, more users per sector (e.g., in MU-MIMO configurations), more antennas per sector, and generally supporting higher density and cheaper operation of the DU function within the network by using more efficient hardware and algorithms to scale more efficiently ¶ [0122]).
Regarding claim 7, the combination of West and O’Shea discloses, wherein the data indicative of the recovered data comprises inferred bits (O’Shea: As another example, known reference or data subcarriers can come from out-of-band coordination from other user equipment (UE), next generation nodeB (gNB) or other base stations, network elements, or prior knowledge of content. In some cases, application data or probabilistic information on one or more of these items can be used to infer transmitted symbols ¶ [0070]. Also see ¶ [0108], [0110], [0121]).
Regarding claim 8, the combination of West and O’Shea discloses, comprising training the first machine learning model and the second machine learning model in a joint training process, wherein training data in the joint training process comprises RF resource grids, and wherein labels for the training data includes ground-truth inferred bits corresponding to the RF resource grids or ground-truth recovered data corresponding to the RF resource grids (O’Shea: In some implementations, a set of profiles, such as urban, rural, indoor, macro, micro, femto, or other profiles related to channel behavior correlated to or predicted by deployment scenario, is used to determine an initial model of the machine-learning network 212 that is deployed, and used to configure processes for determining augmentation or other training parameters for the machine-learning network 212. In some instances, data or models may be shared in cloud environments or network sharing configurations between specific gNB cells to improve initial machine-learning network models, or to jointly improve models within multiple environments with shared phenomena. For example, cells within a grid of cells that share similar interference, cells with similar delay spreads, or cells with other similar behaviors, can be used to improve the effectiveness, speed, or performance of the machine-learning network 212 ¶ [0079]. Also see ¶ [0052], [0053]).
Regarding claim 9, the combination of West and O’Shea discloses, wherein the at least one machine learning model is configured to receive inputs of varying sizes (West: FIG. 6 illustrates an example augmentation routine 600. The augmentation routine 600 creates unique time-series vectors from one or more time series samples 602 for training one or more machine learned models. The augmentation routine 600 increases the size of the training dataset as well as generalizes models to new electromagnetic environments and analog front-ends with different components ¶ [0114]).
Regarding claim 10, the combination of West and O’Shea discloses, wherein the at least one machine learning model comprises a fully convolutional neural network (O’Shea: the machine-learning network is a fully convolutional neural network ¶ [0016]. Also see ¶ [0057]).
Regarding claim 11, the combination of West and O’Shea discloses, wherein the samples of the RF uplink data signals are provided as input in an orthogonal frequency division multiplexing (OFDM) resource grid form (O’Shea: The data signal is generated using a modulator for orthogonal frequency-division multiplexing (OFDM) systems [Abstract]. Also see ¶ [0006], [0007], [0018]).
Regarding claim 12, the combination of West and O’Shea discloses, wherein the samples of the RF uplink data signals comprise a subset of an uplink resource grid, the subset corresponding to an RF signal burst received from a user device (O’Shea: Alternatively or in addition, signals may be augmented by the superposition of multiple signals, the synthesis of new wide-band signal scenarios from bursts either simulated or composed from one or more real signal recordings, the introduction of new propagation effects such as fading patterns, interference effects, mobility models, ray tracing models, or a wide range of other signal processing routines designed to vary the signals in ways that occur randomly in real world systems. In order to select the suitable set of augmentation strategies for one or more radio signal bands while training, random subsets and parameterizations of the augmentation routines may be used. In order to select a target augmentation strategy, a validation loss may be evaluated for a machine learned model trained using the augmented version of a training set ¶ [0126]. Also see ¶ [0145]).
Regarding claim 13, the combination of West and O’Shea discloses, comprising executing the at least one machine learning model in an L1 layer of a distributed unit (DU) (O’Shea: The split in the PHY layer typically concerns allotting processing stages to either a RU or DU within a system ¶ [0112]. Also see ¶ [0113], [0135]).
Regarding claim 14, the combination of West and O’Shea discloses, comprising training a first machine learning model of the at least one machine learning model, wherein the training comprises adjusting weights and parameters of the first machine learning model based on a loss function, wherein the loss function (O’Shea: In some implementations, updating the machine-learning network based on the error term includes determining, based on a loss function, a rate of change of one or more weight values within the machine-learning network; and performing an optimization process using the rate of change to update the one or more weight values within the machine-learning network ¶ [0013], [0068], [0097]) is based on a comparison of (i) pilot values in uplink resource grids and (ii) ground truth values corresponding to the pilot values (O’Shea: The modified pilot and data information are processed using a machine-learning network. A prediction corresponding to the data signal transmitted through the communications channel is obtained from the machine-learning network. The prediction is compared to a set of ground truths and updates, based on a corresponding error term, are applied to the machine-learning network [Abstract]. Also see ¶ [0007], [0012], [0095]).
Regarding claim 15, the combination of West and O’Shea discloses, comprising training a first machine learning model of the at least one machine learning model, wherein the training comprises adjusting weights and parameters of the first machine learning model based on a loss function, wherein the loss function (O’Shea: In some implementations, updating the machine-learning network based on the error term includes determining, based on a loss function, a rate of change of one or more weight values within the machine-learning network; and performing an optimization process using the rate of change to update the one or more weight values within the machine-learning network ¶ [0013], [0068], [0097]) is based on a comparison of (i) data values in uplink resource grids and (ii) ground truth values corresponding to the data values (O’Shea: a number of systems use multi-carrier signal modulation schemes, such as OFDM, to transmit information. Some of the time-frequency subcarriers within an OFDM grid can be allocated as reference tones or pilot signals. Pilot signals can be resource elements with known values; these can be referred to as pilot resource elements. Other resource elements within the OFDM grid can carry data; these can be referred to as data resource elements ¶ [0031], [0052], [0058]. Upon knowledge of a correct frame (e.g., the checksum passes, LDPC check bits are correct), bits can be re-modulated to provide ground truth symbol values, correct bits, or log-likelihood ratios can be computed from the received and ground truth symbol locations or channel estimates. These ground truth symbol values, correct bits, or log-likelihood ratios, among others, can be used within the distance metric in order to update the machine learning model and its weights ¶ [0069], [0095]).
Regarding claim 16, the combination of West and O’Shea discloses, comprising simulating channel effects on the ground truth values, to obtain the data values as simulated pilot values (O’Shea: Implementations may include one or more of the following features. In some implementations, a machine-learning network performs operations corresponding to pilot estimation, interpolation, and equalization. The communications channel may be a simulated channel that includes one or more of an Additive White Gaussian Noise (AWGN) or Rayleigh fading channel model, International Telecommunication Union (ITU) or 3.sup.rd Generation Partnership Project (3GPP) fading channel models, emulated radio emissions, propagation models, ray tracing within simulated geometry or an environment to produce channel effects, or a machine-learning network trained to approximate measurements over a real channel ¶ [0016]. O’Shea: The modified pilot and data information are processed using a machine-learning network. A prediction corresponding to the data signal transmitted through the communications channel is obtained from the machine-learning network. The prediction is compared to a set of ground truths and updates, based on a corresponding error term, are applied to the machine-learning network [Abstract]. Also see ¶ [0007], [0012], [0095]).
Regarding claim 17, the combination of West and O’Shea discloses, wherein the at least one machine learning model is configured to perform patch-based processing of the samples of the RF uplink data signals (O’Shea: see Fig. 1, wherein input and output are patches which is processed by ML model).
Regarding claim 18, the combination of West and O’Shea discloses, wherein the at least one machine learning model has an architecture that includes at least one of: a non-batch norm, or a Smooth ReLU activation function (O’Shea: the machine-learning network 212 includes non-linearity, such as a rectified linear unit (ReLU), sigmoid, parametric rectified linear unit (PReLU) ¶ [0064]).
Conclusion
The examiner requests, in response to this Office action, support be shown for language added to any original claims on amendment and any new claims. That is, indicate support for newly added claim language by specifically pointing to page(s) and line no(s) in the specification and/or drawing figure(s). This will assist the examiner in prosecuting the application.
When responding to this office action, Applicant is advised to clearly point out the patentable novelty which he or she thinks the claims present, in view of the state of the art disclosed by the references cited or the objections made. He or she must also show how the amendments avoid such references or objections See 37 CFR 1.111(c).
Any inquiry concerning this communication or earlier communications from the examiner should be directed to MOHAMMAD S ROSTAMI whose telephone number is (571)270-1980. The examiner can normally be reached Mon-Fri From 9 a.m. to 5 p.m..
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, Boris Gorney can be reached at (571)270-5626. 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.
12/16/2025
/MOHAMMAD S ROSTAMI/ Primary Examiner, Art Unit 2154