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 .
Information Disclosure Statement
The information disclosure statement (IDS) submitted on 06/11/2026 was filed. The submission is in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statement is being considered by the examiner
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.
Claims 1-20 are rejected under 35 U.S.C. 103 as being unpatentable over Lee (US 2021/0110261 A1) in view of 3GPP TS 38.214 version 16.8.0.
Regarding claim 1, Lee discloses: A channel feature information transmission method, comprising: inputting, by a terminal, channel information of each layer into a corresponding first artificial intelligence (AI) network model for processing, and obtaining channel model, because Lee teaches a user equipment that pre-processes an estimated downlink channel matrix and feeds the resulting matrix into a trained transmit-side neural network that converts it into a compact codeword vector representing the channel: (Lee, para [0106] “The UE may perform pre - processing 704 of an estimated downlink channel matrix H 703 so as to produce a new matrix X. The Tx NN 705 may convert the pre - processed output matrix à into a codeword vector.”)
Furthermore, Lee discloses: wherein one layer corresponds to one first AI network model, because Lee teaches an autoencoder arrangement in which a transmit-side network with an input layer and hidden layers is paired with the data it encodes, so that a distinct network instance is associated with the input it processes: (Lee, para [0103] “the autoencoder NN may include a Tx NN 603 including an input layer and an Rx NN 604 including an output layer. The Tx NN 603 may include an input layer and at least one hidden layer”)
Although Lee teaches a user equipment that pre-processes an estimated downlink channel matrix, encodes it with a trained transmit-side neural network into a codeword vector, and feeds that compressed channel representation back to the base station as a CSI report: (Lee, [0103], [0106]), Lee does not explicitly disclose per-layer organization of the reported channel state information, in which the number of spatial transmission layers is set by the reported rank and separate channel state information is derived and reported for each layer.
However, Lee in view of TS 38.214 discloses reporting, by the terminal, the channel feature information corresponding to each layer to a network side device because TS 38.214 teaches channel state information is defined to include a rank indicator together with per-layer precoding information, and the reporting framework expressly carries an indication of coefficients per layer, so the reported quantities are organized on a layer-by-layer basis with one set of reported information for each spatial layer (TS 38.214, cl. 5.2.3, “For Type II CSI feedback, Part 1 contains RI (if reported), CQI, and an indication of the number of non-zero wideband amplitude coefficients per layer for the Type II CSI”; TS 38.214, cl. 5.2.1.4, “The LI indicates which column of the precoder matrix of the reported PMI corresponds to the strongest layer of the codeword corresponding to the largest reported wideband CQI.”).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date to organize the neural-network-encoded channel feedback of Lee on a per-layer basis, and to report the resulting per-layer channel feature information as taught by TS 38.214, because TS 38.214 shows that channel state feedback is routinely structured per spatial layer (rank indicator, per-layer precoder columns, and per-layer coefficient indications), and applying that established per-layer structure to Lee's autoencoder feedback is no more than using a known reporting organization for its established purpose of matching the feedback to the number of spatial streams the channel supports, with the predictable result of more accurate multi-layer channel reconstruction at the base station.
Regarding claim 2, in spite of the fact that Lee teaches neural-network encoding of an estimated downlink channel matrix into a codeword vector that is fed back to the base station as a CSI report: (Lee, para. [0103], [0106]), Lee does not explicitly disclose distinct per-layer models whose output lengths decrease across the layer sequence.
Yet, Lee in view of TS 38.214 discloses The method according to claim 1, wherein first Al network models corresponding to all layers are different, and lengths of channel feature information output by the first AI network models gradually decrease in a sequence of the layers because the reporting framework orders layers by strength, the strongest layer being identified relative to the largest reported wideband CQI, and it carries an explicit per-layer count of non-zero amplitude coefficients, so the amount of information reported varies from layer to layer and decreases for the weaker layers later in the ordering; applying separate encoders per layer with correspondingly shorter outputs for those later layers follows directly from this graded per-layer payload (TS 38.214, cl. 5.2.3, “For Type II CSI feedback, Part 1 contains RI (if reported), CQI, and an indication of the number of non-zero wideband amplitude coefficients per layer for the Type II CSI” … cl. 5.2.1.4, “The LI indicates which column of the precoder matrix of the reported PMI corresponds to the strongest layer of the codeword corresponding to the largest reported wideband CQI.”).
Thus, it would have been obvious to one of ordinary skill in the art before the effective filing date to allocate different per-layer encoders with progressively shorter outputs in the feedback scheme of Lee in the manner taught by TS 38.214, because TS 38.214 shows layers being ranked by strength and a per-layer count of reported coefficients, which teaches the skilled artisan that weaker layers warrant fewer feedback bits, and because Lee itself seeks to transmit accurate channel information using a limited number of feedback bits, so grading the encoder output length per layer is a predictable way to spend the available uplink payload where it contributes most.
Regarding claim 3, although Lee teaches neural-network encoding of an estimated downlink channel matrix into a codeword vector that is fed back to the base station as a CSI report: (Lee, para. [0103], [0106]), Lee does not explicitly disclose determination of the number of layers from the channel rank and selection of a per-layer model according to a target-parameter proportion.
Yet, Lee in view of TS 38.214 discloses The method according to claim 1, determining, by the terminal based on a rank of a channel, a quantity of layers corresponding to the channel information because the reported rank indicator is the quantity that fixes how many spatial layers the feedback covers, and the framework expressly conditions the remaining per-layer quantities (PMI, CQI, LI) on that reported rank, so the terminal first establishes the number of layers from the rank before deriving per-layer feedback (TS 38.214, cl. 5.2.1.4, “- LI shall be calculated conditioned on the reported CQI, PMI, RI and CRI-CQI shall be calculated conditioned on the reported PMI, RI and CRI-PMI shall be calculated conditioned on the reported RI and CRI-RI shall be calculated conditioned on the reported CRI.”).
Moreover, TS 38.214 discloses obtaining, by the terminal, a proportion of a target parameter of a first target layer to a sum of target parameters of second target layers, and determining, based on a proportion range in which the proportion is located, a first AI network model corresponding to the first target layer because the framework compares per-layer quality measures against one another, identifying the strongest layer by reference to the largest reported wideband CQI and resolving ties among equal-valued reports, which is a relative (proportional) evaluation of a per-layer parameter across the set of layers; using the resulting relative ranking to pick the encoding treatment applied to a given layer is the straightforward application of that comparison (TS 38.214, cl. 5.2.1.4, (page 57) “The LI indicates which column of the precoder matrix of the reported PMI corresponds to the strongest layer of the codeword corresponding to the largest reported wideband CQI. If two wideband CQIs are reported and have equal value, the LI corresponds to strongest layer of the first codeword.”).
Furthermore, TS 38.214 discloses wherein different proportion ranges correspond to different first AI network models, and the target parameter comprises any one of the following: an eigenvalue, a channel quality indicator (CQI), or a channel capacity because the channel quality indicator is expressly one of the reported per-layer quantities used to rank layers, and its indices are mapped to defined interpretation ranges, so partitioning the parameter into ranges and associating each range with a different treatment is taught by the reference itself (TS 38.214, cl. 5.2.2.1, “The CQI indices and their interpretations are given in Table 5.2.2.1-2 or Table 5.2.2.1-4 for reporting CQI based on QPSK, 16QAM and 64QAM.”).
Consequently, it would have been obvious to one of ordinary skill in the art before the effective filing date to derive the layer count from the reported rank and to select the per-layer encoder according to the relative magnitude of a per-layer quality measure, as taught by TS 38.214, in the autoencoder feedback of Lee, because TS 38.214 conditions all per-layer feedback on the reported rank and ranks layers by their reported CQI, giving the skilled artisan a concrete and conventional basis for both sizing the feedback to the number of streams and matching encoder capacity to each layer's relative contribution, with the predictable benefit of accurate channel reconstruction within a limited feedback budget.
Regarding claim 4, in spite of the fact that Lee teaches neural-network encoding of an estimated downlink channel matrix into a codeword vector that is fed back to the base station as a CSI report: (Lee, para. [0103], [0106]), Lee does not explicitly disclose distinct per-layer models fed with the channel information of a layer sorted by a target parameter.
Yet, Lee in view of TS 38.214 discloses The method according to claim 1, wherein first AI network models corresponding to all layers are different, and an input of a target first Al network model comprises channel information of a third target layer because per-layer precoder columns are reported separately for each spatial layer, so the channel information belonging to one particular layer is the quantity processed and reported for that layer, and a distinct encoder assigned to each layer takes that layer's own channel information as its input (TS 38.214, cl. 5.2.3, “For Type II CSI feedback, Part 1 contains RI (if reported), CQI, and an indication of the number of non-zero wideband amplitude coefficients per layer for the Type II CSI”).
Moreover, TS 38.214 discloses wherein layers corresponding to the terminal are sorted based on a target parameter, the third target layer is any one of the sorted layers corresponding to the terminal, the target first Al network model is a first Al network model corresponding to the third target layer, and the target parameter comprises any one of the following: an eigenvalue, a CQI, or a channel capacity because layers are expressly ordered by strength, with the strongest layer identified by reference to the largest reported wideband CQI and an explicit tie-break rule, which is a sorting of the terminal's layers by a per-layer parameter; any layer in that sorted order is then processed by the encoder assigned to it (TS 38.214, cl. 5.2.1.4, “The LI indicates which column of the precoder matrix of the reported PMI corresponds to the strongest layer of the codeword corresponding to the largest reported wideband CQI. If two wideband CQIs are reported and have equal value, the LI corresponds to strongest layer of the first codeword.”).
Furthermore, TS 38.214 discloses wherein layers corresponding to the terminal are sorted based on a target parameter, the third target layer is any one of the sorted layers corresponding to the terminal, the target first Al network model is a first Al network model corresponding to the third target layer, and the target parameter comprises any one of the following: an eigenvalue, a CQI, or a channel capacity because the channel quality indicator is one of the enumerated reported quantities of the channel state information, and it is the very measure used to identify the strongest layer, so the sorting parameter recited in the alternative is expressly present (TS 38.214, cl. 5.2.1.1, “CSI may consist of Channel Quality Indicator (CQI), precoding matrix indicator (PMI), CSI-RS resource indicator (CRI), SS/PBCH Block Resource indicator (SSBRI), layer indicator (LI), rank indicator (RI), L1-RSRP or L1-SINR.”).
For these reasons, it would have been obvious to one of ordinary skill in the art before the effective filing date to sort the terminal's spatial layers by a reported quality measure and to assign each sorted layer its own encoder taking that layer's channel information as input, as taught by TS 38.214, within the autoencoder feedback scheme of Lee, because TS 38.214 already ranks layers by reported CQI and reports precoding information separately per layer, so a PHOSITA would have organized Lee's learned encoder along the same per-layer ordering to devote encoding resources to the layers contributing most to throughput, a predictable use of a known feedback structure.
Regarding claim 5, even though Lee teaches neural-network encoding of an estimated downlink channel matrix into a codeword vector that is fed back to the base station as a CSI report, with per-layer sorting and per-layer encoders: (Lee, para. [0103], [0106]), Lee does not explicitly disclose an input to a later layer's model that further comprises information derived from a preceding layer in the sorted order.
Yet, Lee in view of TS 38.214 discloses The method according to claim 4, wherein the third target layer is any one of the sorted layers corresponding to the terminal except the first layer, and an input of the target first Al network model further comprises any one of the following: an output of a first AI network model corresponding to a previous layer of the third target layer because the reported per-layer quantities are expressly computed conditioned on quantities already determined for the report, the strongest (first) layer being identified before the remaining layers are characterized, so information established for an earlier layer in the ordering is carried forward and used when deriving the feedback for a subsequent layer (TS 38.214, cl. 5.2.1.4, (page 52) “- LI shall be calculated conditioned on the reported CQI, PMI, RI and CRI-CQI shall be calculated conditioned on the reported PMI, RI and CRI-PMI shall be calculated conditioned on the reported RI and CRI-RI shall be calculated conditioned on the reported CRI.”).
Moreover, TS 38.214 discloses channel information corresponding to the previous layer of the third target layer because the precoder column identified for the strongest layer is reported and then relied upon when the remaining layers are characterized, so the channel information of the preceding layer in the sorted ordering is available as an input to the derivation performed for the following layer (TS 38.214, cl. 5.2.1.4, “The LI indicates which column of the precoder matrix of the reported PMI corresponds to the strongest layer of the codeword corresponding to the largest reported wideband CQI.”).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date to condition the encoding of a later sorted layer on information already derived for a preceding layer, as taught by TS 38.214, in the autoencoder feedback of Lee, because TS 38.214 expressly derives each per-layer feedback quantity conditioned on quantities already fixed for the report and identifies the strongest layer first, and a PHOSITA applying that sequential dependency to Lee's learned encoder would predictably remove redundancy between correlated layers and thereby reduce the feedback payload, consistent with Lee's stated aim of conveying accurate channel information with a limited number of feedback bits.
Regarding claim 6, which depends on claim 1, Lee discloses The method according to claim 1, inputting, by the terminal, the channel information of each layer into the corresponding first AI network model for processing after preprocessing the channel information of each layer, as Lee further discloses pre-processing of the estimated downlink channel matrix to produce a new matrix, which is then supplied to the transmit-side neural network that converts it into a codeword vector (Lee, para [0106] “The UE may perform pre - processing 704 of an estimated downlink channel matrix H 703 so as to produce a new matrix X. The Tx NN 705 may convert the pre - processed output matrix à into a codeword vector.”).
Regarding claim 7, Lee discloses: The method according to claim 6, wherein the inputting, by the terminal, the channel information of each layer into the corresponding first AI network model for processing after preprocessing the channel information of each layer comprises any one of the following: inputting, by the terminal, the channel information of each layer into the corresponding first Al network model after preprocessing the channel information of each layer by using a second Al network model, because Lee teaches a transmit-side neural network having an input layer and at least one hidden layer, so that the channel matrix is transformed by a preceding learned stage whose result is passed onward to the encoding stage that produces the codeword (Lee, para [0103] “The Tx NN 603 may include an input layer and at least one hidden layer, and the Rx NN 604 may include an output layer and at least one hidden layer.”).
Moreover, Lee discloses: or after preprocessing channel information of a target layer by using a target second AI network model, inputting, by the terminal, an output of the target second Al network model into a first Al network model corresponding to the target layer, since Lee teaches the output of the pre-processing stage being supplied as the input matrix to the transmit-side network, which encodes it into the codeword vector that is reported (Lee, para [0106] “The Tx NN 705 may convert the pre - processed output matrix à into a codeword vector.”).
Regarding claim 8, which depends on claim 1, Lee discloses The method according to claim 1, wherein the reporting, by the terminal, the channel feature information corresponding to each layer to a network side device comprises post-processing, by the terminal, channel feature information corresponding to a target layer, and reporting the post-processed channel feature information to the network side device, as Lee further discloses conversion of the encoder's codeword vector into a transmissible signal form before it is fed back to the base station as a CSI report (Lee, para [0106] “The codeword vector is converted into a signal in a form that is transmissible via a CSI transmitter 706, and may be fed back to the BS ( CSI report).”).
Lee discloses wherein the target layer is at least one layer corresponding to the terminal, as Lee further discloses the encoded channel representation being produced for the channel the terminal has estimated and reported, so at least one such stream of the terminal is the subject of the reporting (Lee, para [0106] “The BS may obtain à by decoding the codeword vector received via a CSI receiver 707 using the Rx NN 708.”).
Regarding claim 9, even though Lee teaches neural-network encoding of the estimated channel matrix into a codeword vector that is converted to a transmissible form and fed back to the base station as a CSI report: (Lee, para. [0103], [0106]), Lee does not explicitly disclose reduction of the reported feature information to a shorter target length and reporting of that length together with the shortened information.
Yet, Lee in view of TS 38.214 discloses post-processing, by the terminal, the channel feature information corresponding to the target layer to obtain channel feature information of a target length because a portion of the reported channel state information is expressly omitted according to a priority order until the report fits the configured code rate, which is a post-processing of the derived feedback that shortens it to a length the uplink allocation can carry (TS 38.214, cl. 5.2.4, (page 104) “If any of the CSI reports consist of two parts, the UE may omit a portion of Part 2 CSI. Omission of Part 2 CSI is according to the priority order shown in Table 5.2.3-1.”).
Moreover, TS 38.214 discloses wherein the target length is less than a length of the channel feature information that is not post-processed because the omission proceeds beginning with the lowest priority level and continues until the resulting code rate meets the configured limit, so the transmitted report is necessarily shorter than the full, unomitted report (TS 38.214, cl. 5.2.4, “Part 2 CSI is omitted beginning with the lowest priority level until the Part 2 CSI code rate is less or equal to the one configured by higher layer parameter maxCodeRate.”).
Furthermore, TS 38.214 discloses reporting, by the terminal, the target length and the channel feature information of the target length to the network side device because the first part of the two-part report has a fixed payload and is used to identify the number of information bits carried in the second part, and it is transmitted in its entirety before the second part, so the terminal signals the length of the variable-length feedback along with that feedback (TS 38.214, cl. 5.2.3, (page 101) “Part 1 has a fixed payload size and is used to identify the number of information bits in Part 2. Part 1 shall be transmitted in its entirety before Part 2.”).
For these reasons, it would have been obvious to one of ordinary skill in the art before the effective filing date to shorten the encoded channel feedback of Lee to a reduced length and to signal that length to the base station in the manner taught by TS 38.214, because TS 38.214 provides an established two-part reporting mechanism in which a fixed-size first part identifies the number of bits in the variable second part and lower-priority content is dropped to meet the configured code rate, and a PHOSITA would apply that mechanism to the learned codeword feedback of Lee so that the receiver can correctly parse a report whose size varies with the available uplink resources.
Regarding claim 10, although Lee teaches neural-network encoding of the estimated channel matrix into a codeword vector that is post-processed and fed back to the base station as a CSI report: (Lee, para. [0103], [0106]), Lee does not explicitly disclose inclusion of the target length in a first part of the channel state information report.
Yet, Lee in view of TS 38.214 discloses The method according to claim 8, wherein the target length is comprised in a first part of channel state information (CSI) in a case that the channel feature information is reported by using the CSI because the first part of the channel state information report has a fixed payload size and expressly serves to identify the number of information bits carried in the second part, so the length of the variable-length feedback is itself conveyed within the first part of the report (TS 38.214, cl. 5.2.3, “Part 1 has a fixed payload size and is used to identify the number of information bits in Part 2. Part 1 shall be transmitted in its entirety before Part 2.”).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date to carry the length of the encoded channel feedback of Lee in the fixed-size first part of the channel state information report as taught by TS 38.214, because TS 38.214 assigns exactly that role to the first part, and a PHOSITA reporting a variable-length learned codeword over the same uplink feedback channel would predictably use the existing fixed-size field to tell the receiver how many bits of encoded channel information follow.
Regarding claim 11, in spite of the fact that Lee teaches neural-network encoding of the estimated channel matrix into a codeword vector fed back to the base station as a CSI report over an uplink control or shared channel: (Lee, para. [0103], [0106]), Lee does not explicitly disclose reporting of per-layer feature information split between a first part and a second part of the channel state information.
Yet, Lee in view of TS 38.214 discloses The method according to claim 1, wherein in a case that the channel feature information is reported by using the CSI, the reporting, by the terminal, the channel feature information corresponding to each layer to a network side device comprises any one of the following: in a case that the layers corresponding to the terminal are sorted based on the target parameter, reporting, by the terminal, the channel feature information corresponding to the first layer to the network side device by using the first part of the CSI, and reporting channel feature information corresponding to other layers except the first layer by using a second part of the CSI because the report is split into two parts in which the first part carries the rank, the resource indicator and the quality indicator for the first codeword while the second part carries the precoding information, the layer indicator and the quality indicator for the second codeword, so information for the leading layer travels in the first part and information for the remaining layers travels in the second part (TS 38.214, cl. 5.2.3, “- For Type I CSI feedback, Part 1 contains RI (if reported), CRI (if reported), CQI for the first codeword (if reported). Part 2 contains PMI (if reported), LI (if reported) and contains the CQI for the second codeword (if reported) when RI is larger than 4.”).
Moreover, TS 38.214 discloses wherein the target parameter comprises any one of the following: an eigenvalue, a CQI, or a channel capacity because the channel quality indicator is one of the enumerated components of the reported channel state information and is the measure by which the strongest layer is identified, so it serves as the parameter on which the layers are sorted (TS 38.214, cl. 5.2.1.4, “The LI indicates which column of the precoder matrix of the reported PMI corresponds to the strongest layer of the codeword corresponding to the largest reported wideband CQI.”).
Accordingly, it would have been obvious to one of ordinary skill in the art before the effective filing date to carry the per-layer encoded channel feedback of Lee across the two-part report structure taught by TS 38.214, placing the leading layer's information in the first part and the remaining layers' information in the second part, because TS 38.214 already partitions channel state feedback that way and ranks layers by their reported quality indicator, and a PHOSITA would predictably adopt that partition so the most critical feedback is protected in the fixed-size part while lower-priority per-layer content can be omitted if the uplink allocation is insufficient.
Regarding claim 12, although Lee teaches neural-network encoding of the estimated channel matrix into a codeword vector fed back to the base station as a CSI report: (Lee, para. [0103], [0106]), Lee does not explicitly disclose discarding of reported feature information in a reverse order of the layer order.
Yet, Lee in view of TS 38.214 discloses The method according to claim 1, wherein the reporting, by the terminal, the channel feature information corresponding to each layer to a network side device comprises: reporting, by the terminal, the channel feature information corresponding to each layer to the network side device, and discarding the channel feature information in a reverse order of an order of the layers because the report is omitted beginning with the lowest priority level and continues until the code rate constraint is met, and because the layers are ordered by strength with the strongest first, dropping from the lowest priority upward discards the per-layer content in the reverse of the layer ordering (TS 38.214, cl. 5.2.3, “CSI report is omitted beginning with the lowest priority level until the CSI report code rate is less or equal to the one configured by the higher layer parameter maxCodeRate.” … cl. 5.2.3, “Part 2 CSI is omitted beginning with the lowest priority level until the Part 2 CSI code rate is less or equal to the one configured by higher layer parameter maxCodeRate.”).
Thus, it would have been obvious to one of ordinary skill in the art before the effective filing date to drop the lowest-priority per-layer content of Lee's encoded feedback first, in the reverse of the layer ordering, as taught by TS 38.214, because TS 38.214 sets out exactly that omission rule for fitting a channel state report to the configured code rate, and applying it to a multi-layer learned codeword report predictably preserves the feedback for the strongest layers, which carry most of the achievable throughput, when the uplink allocation cannot carry the entire report.
Regarding claim 13, in spite of the fact that Lee teaches neural-network encoding of a channel matrix estimated from a reference signal received from the base station, fed back as a CSI report: (Lee, para. [0103], [0106], [0114]), Lee does not explicitly disclose determination of the channel rank from a channel state information reference signal measurement.
Yet, Lee in view of TS 38.214 The method according to claim 1, wherein the method further comprises: determining, by the terminal, the rank of the channel based on a CSI reference signal channel estimation result; and the reporting, by the terminal, the channel feature information corresponding to each layer to a network side device comprises: reporting, by the terminal, a rank indicator (RI) and the channel feature information corresponding to each layer to the network side device because the rank indicator is one of the enumerated components of the channel state information the terminal derives, and the terminal derives those parameters from the channel state information reference signal resources configured for channel measurement, so the rank is determined from a reference-signal-based channel estimate and then reported (TS 38.214, cl. 5.2.1.1, “CSI may consist of Channel Quality Indicator (CQI), precoding matrix indicator (PMI), CSI-RS resource indicator (CRI), SS/PBCH Block Resource indicator (SSBRI), layer indicator (LI), rank indicator (RI), L1-RSRP or L1-SINR.” … cl. 5.2.1.4, “the UE shall derive: - the CSI parameters without averaging two or more instances of any periodic or semi-persistent nzp-CSI-RS- Resources in the corresponding NZP-CSI-RS-ResourceSet for channel measurement”).
Consequently, it would have been obvious to one of ordinary skill in the art before the effective filing date to have the terminal of Lee determine the channel rank from measurements on a channel state information reference signal as taught by TS 38.214, because Lee already estimates the downlink channel from reference signals including a CSI-RS, and TS 38.214 shows the rank indicator being derived from those same reference-signal resources, so combining them is the ordinary use of a known measurement procedure to establish how many spatial layers the encoded feedback must cover.
Regarding claim 14, which depends on claim 1, Lee discloses The method according to claim 1, wherein the channel information is precoding information, as Lee further discloses use of an estimated downlink channel matrix, pre-processed into a new matrix, as the quantity supplied to the encoder, where such a matrix expresses the precoding a receiver prefers for the current channel (Lee, para [0092] “Precoding matrix indicator ( PMI): an indicator associated with a precoding matrix that a UE prefers in the current channel state.”).
Regarding claim 15, Lee discloses: wherein one layer of the terminal corresponds to one first AI network model, and the first AI network model is used to process channel information of a layer that is input by the terminal and output the channel feature information, because Lee teaches a transmit-side neural network at the user equipment that takes the pre-processed channel matrix and converts it into the codeword vector that is fed back: (Lee, para [0106] “a Tx NN 705 of the autoencoder NN, which is learned via deep learning, may be disposed in a UE 701, and an Rx NN 708 may be disposed in a BS 702.”)
Although Lee teaches reception at the base station of a neural-network-encoded codeword vector fed back by the user equipment, which the base station decodes and post-processes to recover the downlink channel matrix: (Lee, [0103], [0106]), Lee does not explicitly disclose per-layer organization of the received channel state information, one set of reported information for each spatial transmission layer.
Conversely, Lee in view of TS 38.214 discloses A channel feature information transmission method, comprising: receiving, by a network side device, channel feature information corresponding to each layer that is reported by a terminal; because TS 38.214 teaches the reported channel state information the network receives is structured per spatial layer, carrying an explicit indication of coefficients for each layer along with the rank, so the network side receives channel information corresponding to each layer reported by the terminal (TS 38.214, cl. 5.2.3, “For Type II CSI feedback, Part 1 contains RI (if reported), CQI, and an indication of the number of non-zero wideband amplitude coefficients per layer for the Type II CSI”).
For these reasons, it would have been obvious to one of ordinary skill in the art before the effective filing date to have the base station of Lee receive the neural-network-encoded channel feedback organized per spatial layer as taught by TS 38.214, because TS 38.214 structures received channel state feedback layer by layer in step with the reported rank, and a PHOSITA seeking to reconstruct a multi-layer downlink channel from Lee's learned codeword would predictably adopt that per-layer structure so the decoder can associate each received codeword with the transmission layer it describes.
Regarding claim 16, the claim recites: The method according to claim 15, wherein first AI network models corresponding to all layers are different, and lengths of channel feature information output by the first Al network models gradually decrease in a sequence of the Claim 16 is analogous to claim 2 and is rejected for the same reasons.
Regarding claim 17, the claim recites: The method according to claim 15, wherein in a case that the channel feature information is reported by using CSI, the receiving, by a network side device, channel feature information corresponding to each layer that is reported by a terminal comprises any one of the following: in a case that the layers corresponding to the terminal are sorted based on a target parameter, receiving, by the network side device, channel feature information corresponding to the first layer that is reported by the terminal by using a first part of the CSI, and channel feature information corresponding to other layers except the first layer that is reported by using a second part of the CSI, wherein the target parameter comprises any one of the following: an eigenvalue, a CQI, or a channel capacity; receiving, by the network side device, channel feature information corresponding to each layer that is reported by the terminal by using the second part of the CSI; or receiving, by the network side device, channel feature information corresponding to each layer that is reported by the terminal by using a corresponding block in the second part of the CSI, wherein one layer corresponds Claim 17 is analogous to claim 11 and is rejected for the same reasons.
Regarding claim 18, even though Lee teaches reception at the base station of a neural-network-encoded codeword vector reported by the terminal, decoded to recover the downlink channel matrix: (Lee, para. [0103], [0106]), Lee does not explicitly disclose reception of a rank indicator together with the per-layer channel feature information.
Yet, Lee in view of TS 38.214 discloses The method according to claim 15, wherein the receiving, by a network side device, channel feature information corresponding to each layer that is reported by a terminal comprises: receiving, by the network side device, a rank indicator (RI) and the channel feature information corresponding to each layer that are reported by the terminal because the first part of the received report expressly contains the rank indicator, and the report also carries per-layer information such as the coefficient indication for each layer, so the network side receives the rank indicator together with information corresponding to each spatial layer in the same feedback (TS 38.214, cl. 5.2.3, “For Type II CSI feedback, Part 1 contains RI (if reported), CQI, and an indication of the number of non-zero wideband amplitude coefficients per layer for the Type II CSI”).
Consequently, it would have been obvious to one of ordinary skill in the art before the effective filing date to have the base station of Lee receive a rank indicator alongside the per-layer encoded channel feedback as taught by TS 38.214, because TS 38.214 places the rank indicator in the fixed-size first part of the report precisely so the receiver knows how many layers of feedback follow, and a PHOSITA decoding Lee's learned codewords would need that count to parse and reconstruct the multi-layer channel correctly.
Regarding claim 19, the claim recites: A terminal, comprising a processor and a memory, wherein the memory stores a program or an instruction that can be run on the processor, wherein the program or the instruction, when executed by the processor, causes the terminal to perform: inputting channel information of each layer into a corresponding first artificial intelligence (AI) network model for processing, and obtaining channel feature information output by the first Al network model, wherein one layer corresponds to one first AI network model; and reporting the channel feature information corresponding Claim 19 is analogous to claim 1 and is rejected for the same reasons.
Regarding claim 20, the claim recites: A network side device, comprising a processor and memory, wherein the memory stores a program or instruction that can be run on the processor, and when the program or the instruction is executed by the processor, the steps of the channel feature information transmission method according to claim 15 are implemented. Claim 20 is analogous to claim 15 and is rejected for the same reasons.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to CHONGSUH (John) PARK whose telephone number is 408-918-7574. The examiner can normally be reached Monday - Friday 8:00-5:30 PST
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, Avellino, Joseph can be reached at 571-272-3905 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.
/CHONGSUH PARK/Examiner, Art Unit 2478
/JOSEPH E AVELLINO/Supervisory Patent Examiner, Art Unit 2478
Prosecution Insights