# US 12,224,885 B1 · Machine learning based channel precoder selection for downlink > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 12,224,885 B1 - Document ID: US12224885B1 - Status: Granted - Kind: B1 - Application number: 18/447,942 - Filed: 2023-08-10 - Published: 2025-02-13 - Granted: 2025-02-11 - Earliest claimed: 2023-08-10 - Assignees: QUALCOMM Incorporated - Applicants: QUALCOMM Incorporated - Inventors: Amit Bar-Or Tillinger; Shay Landis; Idan Michael Horn; Jacob Pick; Yehonatan Dallal - Web page: https://patentgrep.com/patent/US12224885B1 ## CPC - H04L 25/0254: {using neural network algorithms} - H04L 25/0224: {using sounding signals} - H04L 25/03165: {using neural networks} - H04W 28/18: Negotiating wireless communication parameters - H04W 72/1273: of downlink data flows - H04L: TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION (arrangements common to telegraphic and telephonic communication H04M) - H04W: WIRELESS COMMUNICATION NETWORKS (broadcast communication H04H; communication systems using wireless links for non-selective communication, e.g. wireless extensions H04M1/72) ## Abstract Methods, systems, and devices for wireless communications are described. A user equipment (UE) may transmit neural network (NN) capability information to a network entity. The network entity may train the NN model by selecting a NN-based precoder and generating associated coefficients. The network entity may transmit an indication of the precoder coefficients to the UE. The network entity may transmit demodulation reference signals (DMRSs) that have not been precoded and a physical downlink shared channel (PDSCH) signal that has been narrowband-precoded according to the indicated precoder. The UE may then perform and input a channel estimation to the NN model. The NN model may output the narrowband precoder, which the UE may use to generate a narrowband channel estimate and demodulate the narrowband precoded PDSCH signal. In some examples, the network entity may update NN coefficients and may indicate the updated NN coefficients to the UE. ## Claims ### Claim 1 A user equipment (UE), comprising: one or more memories storing processor-executable code; and one or more processors coupled with the one or more memories and individually or collectively operable to execute the code to cause the UE to: transmit a capability message indicating one or more neural network characteristics supported by the UE; receive, based at least in part on transmitting the capability message, a neural network characteristic message indicating one or more neural network coefficients of a neural network model corresponding to precoder calculation based on channel estimation; receive downlink signaling and one or more demodulation reference signals, wherein a narrowband precoder corresponding to the downlink signaling is different than a precoding status corresponding to the one or more demodulation reference signals; and decode the downlink signaling according to an output from the neural network model indicating the narrowband precoder of the downlink signaling, wherein the output from the neural network model is based at least in part on the one or more neural network coefficients and an input to the neural network model comprising one or more channel estimation values. ### Claim 2 (depends on 1) The UE of claim 1, wherein the one or more processors are individually or collectively further operable to execute the code to cause the UE to: transmit, via the capability message, an indication of a quantity of layers of the neural network model, a connectivity value of the neural network model supported by the UE, or a combination thereof. ### Claim 3 (depends on 1) The UE of claim 1, wherein the one or more processors are individually or collectively further operable to execute the code to cause the UE to: transmit, via the capability message, an indication of one or more precoders that the UE supports for decoding the downlink signaling, the one or more precoders comprising the narrowband precoder of the downlink signaling. ### Claim 4 (depends on 1) The UE of claim 1, wherein the one or more processors are individually or collectively further operable to execute the code to cause the UE to: receive control signaling indicating a mapping between the output of the neural network model and one or more precoders comprising the narrowband precoder; and select the narrowband precoder from the one or more precoders according to the output of the neural network model based at least in part on the mapping, wherein decoding the downlink signaling is based at least in part on the selecting. ### Claim 5 (depends on 1) The UE of claim 1, wherein the one or more processors are individually or collectively further operable to execute the code to cause the UE to: receive control signaling indicating a granularity of the narrowband precoder indicated by the output of the neural network model, wherein decoding the downlink signaling is based at least in part on applying the narrowband precoder according to the indicated granularity. ### Claim 6 (depends on 1) The UE of claim 1, wherein the one or more processors are individually or collectively further operable to execute the code to cause the UE to: perform a channel estimate according to the one or more demodulation reference signals; input one or more parameter values corresponding to the channel estimate into the neural network model; calculate the narrowband precoder for the downlink signaling according to the output indicating the narrowband precoder; and demodulate the downlink signaling based at least in part on the channel estimate and the calculated narrowband precoder. ### Claim 7 (depends on 1) The UE of claim 1, wherein the one or more processors are individually or collectively further operable to execute the code to cause the UE to: transmit one or more reference signals; receive, based at least in part on transmitting the one or more reference signals, a second neural network characteristic message indicating one or more updated neural network coefficients of the neural network model; and transmit a feedback message indicating successful receipt of the second neural network characteristic message, wherein receiving the downlink signaling is based at least in part on the one or more updated neural network coefficients of the neural network model. ### Claim 8 (depends on 1) The UE of claim 1, wherein the precoding status of the one or more demodulation reference signals is associated with a wideband precoder that is different than the narrowband precoder. ### Claim 9 (depends on 1) The UE of claim 1, wherein the one or more demodulation reference signals are not precoded according to the precoding status of the one or more demodulation reference signals. ### Claim 10 A network entity, comprising: one or more memories storing processor-executable code; and one or more processors coupled with the one or more memories and individually or collectively operable to execute the code to cause the network entity to: receive a capability message indicating one or more neural network characteristics supported by a user equipment (UE); transmit, based at least in part on receiving the capability message, a neural network characteristic message indicating one or more neural network coefficients of a neural network model corresponding to precoder calculation based on channel estimation; precode downlink signaling according to a narrowband precoder, wherein a precoding status corresponding to one or more demodulation reference signals is different than the narrowband precoder; and transmit the downlink signaling and the one or more demodulation reference signals. ### Claim 11 (depends on 10) The network entity of claim 10, wherein the one or more processors are individually or collectively further operable to execute the code to cause the network entity to: receive, via the capability message, an indication of a quantity of layers of the neural network model, a connectivity value of the neural network model, or a combination thereof, wherein the one or more neural network coefficients are based at least in part on the quantity of layers, the connectivity value, or both. ### Claim 12 (depends on 10) The network entity of claim 10, wherein the one or more processors are individually or collectively further operable to execute the code to cause the network entity to: receive, via the capability message, an indication of one or more precoders that the UE supports for decoding the downlink signaling, the one or more precoders comprising the narrowband precoder of the downlink signaling. ### Claim 13 (depends on 12) The network entity of claim 12, wherein the one or more processors are individually or collectively further operable to execute the code to cause the network entity to: transmit control signaling comprising a set of candidate precoders comprising the one or more precoders, wherein receiving the capability message is based at least in part on transmitting the control signaling comprising the set of candidate precoders. ### Claim 14 (depends on 10) The network entity of claim 10, wherein the one or more processors are individually or collectively further operable to execute the code to cause the network entity to: transmit control signaling indicating a mapping between an output of the neural network model and one or more precoders comprising the narrowband precoder. ### Claim 15 (depends on 10) The network entity of claim 10, wherein the one or more processors are individually or collectively further operable to execute the code to cause the network entity to: transmit control signaling indicating a granularity of the narrowband precoder indicated by an output of the neural network model. ### Claim 16 (depends on 10) The network entity of claim 10, wherein the one or more processors are individually or collectively further operable to execute the code to cause the network entity to: receive one or more reference signals from the UE; perform a channel estimation based at least in part on the one or more reference signals; update the one or more neural network coefficients of the neural network model based at least in part on the channel estimation; transmit a second neural network characteristic message indicating the one or more updated neural network coefficients of the neural network model; and receive a feedback message indicating successful receipt of the second neural network characteristic message. ### Claim 17 (depends on 10) The network entity of claim 10, wherein the precoding status of the one or more demodulation reference signals is associated with a wideband precoder that is different than the narrowband precoder. ### Claim 18 (depends on 10) The network entity of claim 10, wherein the one or more demodulation reference signals are not precoded according to the precoding status of the one or more demodulation reference signals. ### Claim 19 A method for wireless communications at a user equipment (UE), comprising: transmitting a capability message indicating one or more neural network characteristics supported by the UE; receiving, based at least in part on transmitting the capability message, a neural network characteristic message indicating one or more neural network coefficients of a neural network model corresponding to precoder calculation based on channel estimation; receiving downlink signaling and one or more demodulation reference signals, wherein a narrowband precoder corresponding to the downlink signaling is different than a precoding status corresponding to the one or more demodulation reference signals; and decoding the downlink signaling according to an output from the neural network model indicating the narrowband precoder of the downlink signaling, wherein the output from the neural network model is based at least in part on the one or more neural network coefficients and an input to the neural network model comprising one or more channel estimation values. ### Claim 20 (depends on 19) The method of claim 19, further comprising: transmitting, via the capability message, an indication of a quantity of layers of the neural network model, a connectivity value of the neural network model supported by the UE, or a combination thereof. ### Claim 21 (depends on 19) The method of claim 19, further comprising: transmitting, via the capability message, an indication of one or more precoders that the UE supports for decoding the downlink signaling, the one or more precoders comprising the narrowband precoder of the downlink signaling. ### Claim 22 (depends on 19) The method of claim 19, further comprising: receiving control signaling indicating a mapping between the output of the neural network model and one or more precoders comprising the narrowband precoder; and selecting the narrowband precoder from the one or more precoders according to the output of the neural network model based at least in part on the mapping, wherein decoding the downlink signaling is based at least in part on the selecting. ### Claim 23 (depends on 19) The method of claim 19, further comprising: receiving control signaling indicating a granularity of the narrowband precoder indicated by the output of the neural network model, wherein decoding the downlink signaling is based at least in part on applying the narrowband precoder according to the indicated granularity. ### Claim 24 (depends on 19) The method of claim 19, further comprising: performing a channel estimate according to the one or more demodulation reference signals; inputting one or more parameter values corresponding to the channel estimate into the neural network model; calculating the narrowband precoder for the downlink signaling according to the output indicating the narrowband precoder; and demodulating the downlink signaling based at least in part on the channel estimate and the calculated narrowband precoder. ### Claim 25 (depends on 19) The method of claim 19, further comprising: transmitting one or more reference signals; receiving, based at least in part on transmitting the one or more reference signals, a second neural network characteristic message indicating one or more updated neural network coefficients of the neural network model; and transmitting a feedback message indicating successful receipt of the second neural network characteristic message, wherein receiving the downlink signaling is based at least in part on the one or more updated neural network coefficients of the neural network model. ### Claim 26 (depends on 19) The method of claim 19, wherein the precoding status of the one or more demodulation reference signals is associated with a wideband precoder that is different than the narrowband precoder. ### Claim 27 (depends on 19) The method of claim 19, wherein the one or more demodulation reference signals are not precoded according to the precoding status of the one or more demodulation reference signals. ### Claim 28 A method for wireless communications at a network entity, comprising: receiving a capability message indicating one or more neural network characteristics supported by a user equipment (UE); transmitting, based at least in part on receiving the capability message, a neural network characteristic message indicating one or more neural network coefficients of a neural network model corresponding to precoder calculation based on channel estimation; precoding downlink signaling according to a narrowband precoder, wherein a precoding status corresponding to one or more demodulation reference signals is different than the narrowband precoder; and transmitting the downlink signaling and the one or more demodulation reference signals. ### Claim 29 (depends on 28) The method of claim 28, further comprising: receiving, via the capability message, an indication of a quantity of layers of the neural network model, a connectivity value of the neural network model, or a combination thereof, wherein the one or more neural network coefficients are based at least in part on the quantity of layers, the connectivity value, or both. ### Claim 30 (depends on 28) The method of claim 28, further comprising: receiving, via the capability message, an indication of one or more precoders that the UE supports for decoding the downlink signaling, the one or more precoders comprising the narrowband precoder of the downlink signaling.