Communication methods and communication devices

Abstract

A communication method for a receiving end includes receiving a received signal and obtaining information of an original signal based on the received signal. A transmitting end obtains a transmitted signal based on the original signal, transmits the transmitted signal, and the transmitted signal is transformed into the received signal through a channel. A structural causal model is used to associate the transmitted signal and the received signal, wherein multiple causal variables of a causal graph of the structural causal model and a causal structure of the causal graph are determined together.

Description

Technical Field

[0001] This invention relates to a communication method and a communication device, and particularly to a communication method and a communication device that can reduce the transmission error rate. Background Technology

[0002] During wireless communication, signals pass through media such as air (called channels), which may introduce various noises or cause interference, resulting in distortion. Therefore, the received signal cannot be the same as the transmitted signal. To correctly decode the received signal, it is necessary to remove the distortion and noise caused by the channels to obtain the original signal.

[0003] In 5G user scenarios, at the user end, to recover the original signal from the signal transmitted by the radio unit, channel estimation is required, using a channel matrix based on statistical correlation (e.g., the value of each element of the channel matrix calculated by sampling a reference signal at certain time frequencies). Since statistical correlation does not imply causation, using a channel matrix to correlate the received and transmitted signals is inaccurate. Furthermore, the need for channel estimation at the user end consumes significant design resources, power, and CPU utilization. Therefore, existing methods for obtaining the original signal need improvement to reduce transmission error rates. Summary of the Invention

[0004] Therefore, the present invention mainly provides a communication method and a communication device to reduce the transmission error rate.

[0005] This invention discloses a communication method for a receiving end, comprising receiving a received signal, wherein a transmitting end obtains a transmission signal based on an original signal, the transmitting end transmits the transmission signal, the transmission signal is transformed into the received signal through a channel; and obtaining information of the original signal based on the received signal, wherein a structural causal model is used to associate the transmission signal and the received signal, and multiple causal variables of a causal graph of the structural causal model and a causal structure of the causal graph are determined together.

[0006] This invention discloses a communication device for a receiving end, comprising a storage circuit for storing an instruction, the instruction including receiving a received signal, wherein a transmitting end obtains a transmission signal based on an original signal, the transmitting end transmits the transmission signal, and the transmission signal is transformed into the received signal through a channel; and obtaining information of the original signal based on the received signal, wherein a structural causal model is used to associate the transmission signal and the received signal, and a plurality of causal variables of a causal graph of the structural causal model and a causal structure of the causal graph are determined together; and a processing circuit coupled to the storage circuit for executing the instruction stored in the storage circuit.

[0007] This invention discloses a communication method for a transmitting end, comprising: obtaining a transmission signal based on an original signal; transmitting the transmission signal, wherein the transmission signal is transformed into a received signal through a channel; a receiving end obtaining information of the original signal based on the received signal; and using a structural causal model to associate the transmission signal and the received signal, wherein multiple causal variables of a causal graph of the structural causal model and a causal structure of the causal graph are determined together.

[0008] This invention discloses a communication device for a transmitting end, comprising a storage circuit for storing an instruction, the instruction comprising: obtaining a transmission signal based on an original signal; transmitting the transmission signal, wherein the transmission signal is transformed into a received signal through a channel; a receiving end obtaining information of the original signal based on the received signal; using a structural causal model to associate the transmission signal and the received signal, wherein multiple causal variables of a causal graph of the structural causal model and a causal structure of the causal graph are determined together; and a processing circuit coupled to the storage circuit for executing the instruction stored in the storage circuit. Attached Figure Description

[0009] Figure 1 This is a schematic diagram of the communication system according to Embodiment 1 of the present invention.

[0010] Figure 2 and Figure 3 These are schematic diagrams of the causal model of Embodiment 1 of the present invention.

[0011] Figure 4 This is a schematic diagram of the basic data and a partial cause-effect graph of Embodiment 1 of the present invention.

[0012] Figure 5 This is a schematic diagram of causal variables and data in an embodiment of the present invention.

[0013] Figure 6 This is a schematic diagram of a simulated deep learning model.

[0014] Figure 7This is a schematic diagram of the communication system according to Embodiment 1 of the present invention.

[0015] Figure 8 and Figure 9 These are schematic diagrams illustrating the communication method according to an embodiment of the present invention.

[0016] [List of Labels in the Attached Image]

[0017] 10,70: Communication System

[0018] 120: Sending end

[0019] 140: Channel

[0020] 160: Receiver

[0021] 20: Causal Model

[0022] 40g: Basic data

[0023] 60: Simulated Deep Learning Models

[0024] 80,90: Communication method

[0025] CG: Cause and Effect Diagram

[0026] cv (i-1) ,cv i ,cv (j-1) ,cv j ,cv y ,CV1~CV n Causal variables

[0027] CV'1~CV' n Output value

[0028] f (i-1) ,f i ,f (j-1) ,f j Observation function

[0029] N: Noise

[0030] n1~n k : Components

[0031] p(s,v),p ⊥ (s,v),p ～ (s,v): Prior

[0032] p(w x |s,v),p(cv y |s): Causal mechanism

[0033] q(s,v|w x Inference Model

[0034] s: semantic factors

[0035] S800~S806, S900~S906: Steps

[0036] v: Factors of change

[0037] w (i-1) ,w i ,w (j-1) ,w j ,w x :data

[0038] X,X * Transmitted signals

[0039] x1~x m ,x * 1~x * m Transmitted signal components

[0040] Y,Y * Received signal

[0041] y1~y k ,y * 1~y * k Received signal components

[0042] Z,Z * Original signal Detailed Implementation

[0043] Figure 1 This is a schematic diagram of a communication system 10 according to an embodiment of the present invention. The communication system 10 may include a transmitting end 120 and a receiving end 160.

[0044] In one embodiment, the transmitting end 120 may transmit a raw signal Z as a transmitted signal X. In other words, the raw signal Z and the transmitted signal X are substantially the same (e.g., Z = X). The raw signal Z may include transmitted signal components x1 to x2. m Then the transmitted signal X may include transmitted signal components x1 to x2. m Where m is a positive integer. Each transmitted signal component x1 to x2... m It can be transmitted by one of the multiple transmitting antennas of the transmitter 120, through a medium such as air (referred to as channel 140). The transmitted signal X (which may contain data, (in some embodiments) a reference signal or other signal) can be modified (distorted) into the received signal Y through channel 140.

[0045] Receiver 160 can receive signal Y. Received signal Y may include received signal components y1 to y2. kWhere k is a positive integer, and each received signal component y1 to y2 is a positive integer. k It can be received by one of the multiple receiving antennas of the receiver 160. The receiver 160 can reconstruct the received signal Y into a signal that is equivalent to or approximates the original signal Z.

[0046] In one embodiment, the transmitted signal X, the original signal Z, and the received signal Y can satisfy Y = HX + N or Y = HZ + N, where H is the channel matrix of channel 140, which can be an m×k matrix, and N is noise, which can include components n1 to n2. k In other words, the received signal Y received by receiver 160 through the receiving antenna may include the transmitted signal X distorted by the channel matrix H and noise N. The transmitted signal X and the received signal Y are correlated using the channel matrix H. Receiver 160 can determine the correlation between X and H based on X = H. -1 (YN) or Z = H -1 (YN) Remove the distortion and noise N caused by channel 140 to obtain the transmitted signal X (i.e., obtain the original signal Z).

[0047] In one embodiment, the channel matrix H can be replaced with a causal model (e.g., a structural causal model, SCM) to obtain the transmitted signal X (i.e., the original signal Z). Where the transmitted signal X and the received signal Y can be correlated using a causal model, (receiver 160) can infer the transmitted signal X that led to the received signal Y, at least based on abductive reasoning (i.e., backward reasoning), and the received signal Y can be reconstructed into a signal equivalent to or approximating the transmitted signal X. This improves prediction accuracy and reduces the number of training samples required.

[0048] For example, Figure 2 and Figure 3 These are schematic diagrams of the causal model 20 according to Embodiment 1 of the present invention. The causal model 20 is implemented or realized by one of the transmitting end 120, the receiving end 160, and a distributed unit (DU). Figure 2 As shown in (a), the causal model 20 can replace the calculation of the channel matrix H or the deep learning model, thereby enabling the transmitted signal X to be inferred from the received signal Y (i.e., the original signal Z to be determined) based on abductive reasoning. Figure 2 (b) and Figure 3 As shown, the causal variable CV1 can represent the transmitted signal X (for example, the attributes of the causal variable CV1 correspond to the transmitted signal X). Causal variable CVn This can represent the predicted received signal Y, where n is a positive integer. Causal variables CV1~CV n The causal graph used to construct causal model 20. N can represent noise.

[0049] In other words, the transmitted signal X and the received signal Y can be correlated using causal model 20, and the communication system 10 performs encoding, decoding, or channel estimation based on causal model 20. Abductive reasoning allows the received signal Y (as a result) to be inferred as the transmitted signal (as a premise). (Receiver 160) can, at least based on abductive reasoning, deduce the transmitted signal X that led to the received signal Y, and can reconstruct the received signal Y into a signal equivalent to or approximating the transmitted signal X. Although there are many possibilities for the transmitted signal that led to the received signal Y, based on abductive reasoning, the causal graph used by causal model 20 helps to infer the received signal Y as one or more possible transmitted signals X.

[0050] In one embodiment, given that multiple combinations (X, N) of the transmitted signal (can be referred to as the candidate transmitted signal) and noise N that could lead to the received signal Y can be found through inference, the present invention can select at least one combination (X, N) of the transmitted signal (can be referred to as the preferred transmitted signal) and noise N most likely to lead to the received signal Y from the Pareto set based on the estimated signal-to-noise ratio (SNR) (or other algorithm). For example, in the case of a SNR of 64, a transmitted signal that is approximately 64 times (or 10 to 100 times) the noise N can be selected as the transmitted signal X. In the case of a SNR of 8, a transmitted signal that is approximately 8 times (or 1 to 10 times) the noise N can be selected as the transmitted signal X. Channel estimation in this manner, through causal model 20, can reduce the error rate of transmission between the transmitter 120 and the receiver 160.

[0051] In one embodiment, the optimized causal model 20 can be selected based on maximum a posteriori (MAP) and point estimates. Accordingly, the causal variables (e.g., the number of causal variables, the attributes of a causal variable, and the number of attributes of a causal variable) and the causal structure (e.g., the connection between attributes) of the causal graph used by the causal model 20 are determined simultaneously, thus avoiding the problems caused by determining the causal variables first and then the causal structure.

[0052] For example, Figure 4This is a schematic diagram of basic data 40g and a local causal graph CG according to Embodiment 1 of the present invention, wherein (a) and (b) show two possibilities for basic data 40g and causal graph CG, respectively. The causal graph CG can be used as the causal graph for causal model 20.

[0053] In one embodiment, the basic data 40g can be obtained from the space of all observable samples, and therefore can also be referred to as observation data. In one embodiment, the basic data 40g can be obtained from all collected data. In one embodiment, the basic data 40g can be the solution space (i.e., observation data) of the feature vectors of the received or transmitted signals. In one embodiment, the basic data 40g can include or relate to all signals (e.g., received signal Y) received by the transmitter 120 or the receiver 160 at all times and in all ways, or signals (e.g., transmitted signal X). In one embodiment, Figure 2 The state or attribute of causal variable CV1 (at all times) can serve as at least part of the underlying data 40g for causal variable CV2, and so on. Figure 2 Causal variables CV (n-1) The state or property (at all times) can be used as a causal variable CV n At least part of the basic data is 40g.

[0054] exist Figure 4 The causal structure of causal graph CG can provide causal variables (e.g., causal variable cv). (i-1) ,cv i ,cv (j-1) ,cv j The relationship between ) and the observation function f (i-1) f i f (j-1) f j Then the basic data 40g of data can be used as w (i-1) w i w (j-1) w j Mapping to causal variable cv (i-1) ,cv i ,cv (j-1) ,cv j And providing causal variable cv (i-1) ,cv i ,cv (j-1) ,cv j Compared to the base data of 40g, the data w (i-1) w i w (j-1) w jThe relationship between them. Here, i and j are positive integers, and the mapping is based on the corresponding data w. i ( Figure 4 (The framed area) instead of the entire base data 40g.

[0055] In one embodiment, the base data 40g of data can be used as w i Assigned to the observation function f i And the posterior probability P(f) of the causal structure C of the causal graph CG. i Maximize C|w) to obtain data from the base data 40g. i Obtain the causal structure C and its causal variable cv. i Therefore, a Bayesian network can be used in conjunction with an observation function (e.g., observation function f). (i-1) f i f (j-1) f j This is used to describe the inference of a causal model. It is worth noting that causal variables (e.g., causal variable cv) are used to describe the inference of a causal model. (i-1) ,cv i ,cv (j-1) ,cv j The causal structure and the causal variables are obtained together, therefore the causal variables (e.g., causal variable cv) are obtained together. (i-1) ,cv i ,cv (j-1) ,cv j The causal structure and the causal structure can influence and constrain each other.

[0056] In one embodiment, the posterior probability P(f) i ,C|w i According to Bayesian rule, P(f) can satisfy the condition. i ,C|w i ,Int)∝P(f i ,C)P(w i |f i (C, Int), where f i C can represent the observation function, and w can represent the causal structure. i This can represent a portion of the baseline data 40g, where Int can represent intervention. In one embodiment, the posterior probability P(f) i ,C|w i It can be proportional to P(f) i ,C)P(w i |f i C) or Where s t-1This can represent the state at time point t-1, where T can represent the current time point, and γ can be 0.5 but is not limited to this. In one embodiment, P(w|f i C) can be In one embodiment, P(w) i，t |s t-1 C, f i ) can be or Where s i,t It can represent a causal variable cv i At a given time point t, Ncv can represent all causal variables (e.g., causal variable cv). (i-1) ,cv i ,cv (j-1) ,cv j The total number of () , where Ncv is a positive integer. This can represent the basic data 40g and the causal variable cv i state s i Compatible data w i The amount of data. In one embodiment, minimizing the amount of data can be utilized. To select the causal variable cv i This allows for the inclusion of frequently used data in the 40g base data set (such as data w). i (Compared to less frequently used data) are divided into smaller parts.

[0057] As can be seen from the above, the Bayesian probability mechanism can combine causal variables (e.g., causal variable cv). (i-1) ,cv i ,cv (j-1) ,cv j The number of causal variables, the state of causal variables, the causal structure, and the observation function of causal variables (e.g., observation function f). (i-1) f i f (j-1) f j This leads to related joint inferences that explain the underlying data 40g, thus generating the causal graph CG. The causal variables in the causal graph CG (e.g., causal variable cv) (i-1) ,cv i ,cv (j-1) ,cv j The number of causal variables (or the number of causal variables) and the causal structure C are determined simultaneously. Based on this, the causal programming module 110P can distinguish... Figure 4 The difference between (a) and (b).

[0058] like Figure 4 As shown, each causal variable (e.g., causal variable cv) iEach of these will correspond to an observation function (e.g., observation function f). i In one embodiment, a causal semantic generative (CSG) model can be used to obtain the observation function (e.g., the observation function f). i This allows for the prediction of low-dimensional state attributes (such as causal variables cv) from high-dimensional environmental variables (such as the base data 40g). i The attributes of the state. If the causal variable (e.g., causal variable cv) (i-1) ,cv i ,cv (j-1) ,cv j ) is manually defined (e.g., defined by a domain expert), and each causal variable (e.g., causal variable cv) is... i Each has a dedicated causal semantic generation observation function CSG() (as the observation function) that makes the causal variables based on the corresponding data (e.g., data w). i ()( Figure 4 (The bounding box area). Furthermore, causal semantic generation models can avoid misclassifying variation factors as causal variables (e.g., causal variable cv). i The cause is the semantic factor, which can be correctly identified as a causal variable (e.g., causal variable cv). i The cause of the causal semantic generation model is determined by the causal invariance principle and involves variational Bayes. In one embodiment, the variable factors and semantic factors may be derived from observed data. In one embodiment, the causal semantic generation model is primarily based on the causal invariance principle and involves variational Bayes.

[0059] In one embodiment, observe the function f i It can satisfy s i,t =f i (w i,t In one embodiment, observe the function f. i This can be achieved using a multivariate Gaussian distribution, for example, one that satisfies... Alternatively, observe the function f i Related to Where z is the causal variable cv in the basic data 40g. i Data that did not contribute, mean μ v Fixed as a zero vector, Σ can be parameterized through Cholesky decomposition, and for example, it can satisfy Σ = LL. TMatrix L can be a lower-triangular matrix with positive diagonals and, for example, can be parameterized to satisfy... matrix L zz It can be a smaller lower triangular matrix, matrix It can be any matrix. L zz It can be parameterized using the sum of the diagonal elements (as confirmed by the exponential map) and the lower triangular matrix (which does not have diagonal elements).

[0060] In one embodiment, causal variables (e.g., causal variable cv) i ) and data (e.g., data w) i The relationship between ) is unknown, but causal variables can be predicted from the data using causal semantic generation models. For example, Figure 5 cv is the causal variable in Embodiment 1 of the present invention. y And a data w x The diagram shows the possible structures of the causal semantic generation model, where (a), (b), (c), and (d) represent the semantic factors, v represents the variable factors, and the solid arrows represent the causal mechanism p(w). x |s,v) and p(cv) y |s), where the dashed arrow represents the inference model q(s,v|w) used for learning. x ).exist Figure 5 In (a), the solid undirected line segment between the semantic factor s and the variable factor v can represent the domain-specific prior p(s,v). Compared to Figure 5 (a) The undirected line segment between semantic factor s and variable factor v. Figure 5 (b) Introducing an independent prior p ⊥ (s,v) := p(s)p(v) reflects the intervention to improve out-of-distribution generalization performance. Compared to Figure 5 (a) The undirected line segment between semantic factor s and variable factor v. Figure 5 (c) Based on the causal invariance principle, a priori p of the point-to-line relationship between semantic factor s and variable factor v is introduced. ～ (s,v) is used to reflect the intervention, thus utilizing unsupervised data. In one embodiment, the causal semantic generation model p:= can be maximized by maximizing the likelihood ratio. <p(s,v),p(wx |s,v),p(cv y The data is fitted to the data and can be calculated using variational inference and the evidence lower bound (ELBO). Then, after applying the reparameterization technique, the expectation can be estimated using Monte Carlo.

[0061] In another embodiment, a deep learning model can be used to obtain the observation function (e.g., the observation function f). i This allows for the prediction of low-dimensional state attributes (such as causal variables cv) from high-dimensional environmental variables (such as the base data 40g). i (attributes of state), however, deep learning models may mistakenly classify changing factors as causal variables (cv) in addition to semantic factors. i The reason.

[0062] exist Figure 2 Because deep learning models require an optimizer to evaluate the inputs given the predicted output, which is very time-consuming and resource-intensive, Figure 2 This approach avoids using deep learning models, thereby improving efficiency and reducing power consumption. Furthermore, by utilizing causal relationships rather than statistical correlations, deep learning models can accurately correlate received signals with transmitted signals, thus reducing transmission error rates.

[0063] In another embodiment, a simulated deep learning model can be used instead of the channel matrix H to obtain the transmitted signal X (i.e., the original signal Z). The transmitted signal X and the received signal Y can be correlated using the simulated deep learning model, and the communication system performs encoding, decoding, or channel estimation based on the simulated deep learning model. For example, Figure 6 This is a schematic diagram of a simulated deep learning model 60. CV'1 can represent the output value of the input layer of the simulated deep learning model 60; for example, CV'1 can represent the transmitted signal X. CV'2~CV' (n-1) This can represent the output value of the hidden layer of a simulated deep learning model 60 after passing through the activation function ReLU(). CV' n This can represent the output value of the output layer of the simulated deep learning model 60, CV' n For example, Y can represent the predicted received signal. N can represent noise.

[0064] Please refer to Figure 2 and Figure 6In one embodiment, a causal graph (e.g., causal graph CG or the causal graph used by causal model 20) can be derived from base data (e.g., base data 40g) based on a simulated deep learning model 60. Figure 2 As shown, the causal graph used in causal model 20 includes causal variables CV1 to CV2. n The causal variable CV1 can be mapped to the output value CV'1 of the input layer of the simulated deep learning model 60. Causal variables CV2 to CV... (n-1) These values ​​can be respectively assigned to the output values ​​CV'2 to CV' of the hidden layer of the simulated deep learning model 60 after the activation function. (n-1) For example, it can satisfy CV'2 = ReLU(CSG(CV'1)) (that is, the output value CV'2 is at least related to the result of the causal semantic generation observation function CSG() mapped to the output value CV'1 using the ReLU() function), CV'3 = ReLU(CSG(ReLU(CSG(CV'1)))), or CV' (n-1) =ReLu(CSG(...ReLu(CSG(CV'1)))), where CSG() can represent the causal semantic generation observation function. Each attribute of the causal variable (e.g., causal variable CV2) can correspond to a neuron in a hidden layer. Causal variable CV n This corresponds to the output value CV' of the output layer of the simulated deep learning model 60. n For example, CV' can be satisfied. n =CSG(CV') (n-l) )+N.

[0065] In one embodiment, in Figure 1 In the process of the transmitted signal X being transformed into the received signal Y through channel 140, it is not very meaningful to investigate the exact value of the noise. Because the noise is randomly varying (e.g., Gaussian distributed), the noise value affecting the reference signal (of the transmitted signal X) is not necessarily the same as the noise value affecting the non-reference signal (e.g., data, control signal, or other signal). Therefore, the noise value decoded from the reference signal (of the received signal Y) is not necessarily the same as the noise value used to decode the non-reference signal (e.g., data, control signal, or other signal). Furthermore, the noise can vary over time or space. However, the overall properties of the randomly varying noise (e.g., the mean, variance, statistical distribution, or other statistical characteristics of the noise) can be used as information for signal-to-noise ratio estimation (SNRestimation). In one embodiment, the noise values ​​(e.g., components n1 to n2) of the noise N in the causal model (e.g., causal model 20) can be used as information for signal-to-noise ratio estimation. k The value of is set to the mean of the noise N.

[0066] There are various methods for estimating the signal-to-noise ratio (SNR). In one embodiment, the average SNR ρ of the received k'-th Orthogonal Frequency-Division Multiplexing (OFDM) preamble is... av Satisfy Where NN' represents a preamble containing NN' modulated subcarriers, S represents the power of the transmitted signal X, C(k', n') represents the data symbol of the n'th subcarrier of the k'th preamble, H(n') represents the channel frequency response, W represents the power of the noise N, η(k', n') represents additive white Gaussian noise (AWGN) with unit variance sampled multiple zero-mean samples, and E() represents the desired value. In one embodiment, the average signal-to-noise ratio ρ of the n'th subcarrier can satisfy... However, the present invention is not limited thereto.

[0067] In one embodiment, the causal graph of the present invention can be verified or processed according to a structural causal model (e.g., Figure 4 Causal diagram (CG).

[0068] Figure 7 This is a schematic diagram of a communication system 70 according to an embodiment of the present invention. The communication system 70 may include a transmitting end 120 and a receiving end 160.

[0069] In one embodiment, the transmitting end 120 can transmit the original signal Z * Convert transmission signal X * To send. Original signal Z * It may be different from the transmitted signal X * For example, the original signal Z * With transmitted signal X * Can satisfy X * =H -1 (Z * -N) or Z * =HX * +N, where H is the channel matrix of channel 140, and the channel matrix H can be an m×k matrix, and N is noise. Transmitted signal X * It may include the transmitted signal component x * 1~x * m Each transmitted signal component x * 1~x * m It can be transmitted by one of the multiple transmitting antennas of the transmitter 120, through channel 140.

[0070] Transmitted signal X * (It may contain data, (in some embodiments) reference signals or other signals) which are altered (distorted) through channel 140 to become the received signal Y. * In one embodiment, the transmitted signal X * Original signal Z * With the received signal Y * It can satisfy Y * =HX * +N=H(H -1 (Z * -N))+N=Z * .

[0071] Receiver 160 can receive signal Y * Received signal Y * It may include the received signal component y * 1~y * k Each received signal component y * 1~y * k The signal can be received by one of the multiple receiving antennas of the receiver 160. The received signal Y received by the receiver 160 through the receiving antenna... * This may include the transmitted signal X after being distorted by the channel matrix H. * and noise N, however, the original signal Z * With the received signal Y * Essentially the same (e.g., Z) * =Y * The original signal Z may include a component y. * 1~y * k The receiver 160 can receive the signal Y. * We obtain the equivalent of the original signal Z * Or approximately the original signal Z * The information. In other words, receiver 160 can directly obtain the original signal Z. * The information received by the receiver 160 does not require X. * =H -1 (Y * -N) to additionally remove distortion and noise caused by channel 140 to restore the original signal Z. * This simplifies the design of the receiver 160 and saves on the computation and power consumption of the central processing unit (CPU).

[0072] In one embodiment, a transmission signal X is transmitted at the transmitting end 120. * Before reaching the receiving end 160, the original signal Z can be altered (distorted).* For transmitting signal X * In this way, receiver 160 can obtain the same signal Z as the original signal without performing channel estimation. * Or approximately the original signal Z * The information. In one embodiment, the mean of the statistical properties of the noise can be used to alter (distort) the original signal Z. * For transmitting signal X * .

[0073] In one embodiment, similar to Figure 1 The communication system 10 can use causal models (e.g., structural causal models, Figure 2 and Figure 3 The causal model 20) replaces the channel matrix H, thus enabling abductive reasoning from the original signal Z. * The transmission signal X is then retracted. In the transmission signal X... * With the received signal Y * When a causal model can be used to correlate the signals, (sender 120) can infer the cause of the received signal Y at least based on abductive reasoning. * (i.e., the original signal Z) * The transmitted signal X * This allows the received signal Y, which is affected by channel 140, to be... * Equivalent to the original signal Z * Or approximately the original signal Z * The signal. Based on abductive reasoning, causal models (e.g.) Figure 2 and Figure 3 The causal model 20) uses a causal graph (e.g. Figure 4 The causal graph (CG) helps to transform the original signal Z * Inferred to be one or more possible transmitted signals X * This improves prediction accuracy and reduces the number of training samples required.

[0074] As can be seen from the above, when transmitting signal X... * At that time, the transmitter 120 can send from the original signal Z * (i.e., receiving signal Y) * The transmitted signal X obtained through abductive reasoning (information) * And the noise N is set to the mean value. The received signal Y received by receiver 160 * Essentially equivalent to the original signal Z * Or approximately the original signal Z * This information eliminates the need for channel estimation, resulting in significant savings in design, power consumption, and CPU utilization. In other words, compared to... Figure 1The communication system 10 and communication system 70 help reduce the power consumption of the receiver 160 and the utilization rate of the central processing unit.

[0075] In another embodiment, similar to Figure 1 The communication system 10 can use simulated deep learning models (e.g.) Figure 6 The simulated deep learning model 60) replaces the channel matrix H.

[0076] In one embodiment, the causal model (e.g., causal model 20) may be stored in the high-level physical layer (H-PHY) of a distributed unit, but is not limited thereto.

[0077] In one embodiment, the original signal Z, the transmitted signal X, and the received signal Y can be radio-frequency signals, and the spectrum can be between 1 kHz and 300 GHz, but is not limited thereto.

[0078] In one embodiment, the communication system 10 can be used for downlink. The transmitting end 120 can be a radio unit (RU) and the receiving end 160 can be customer-premises equipment (CPE), but the invention is not limited thereto.

[0079] For example, the transmitter 120 can be a base station, such as a node B, an evolved-node B (eNB), a next-generation-node B (gNB), a sector, a base transceiver system (BTS), an access point (AP), a relay node, a remote radio head (RRH), a small cell, a base station controller (BSC), or other fixed stations that exchange data and control information with a user terminal or another base station. The receiver 160 can be a user terminal, such as user equipment (UE), terminal equipment, a mobile station (MS), or other fixed or mobile devices.

[0080] In one embodiment, the communication system 10 can be used for uplink. The transmitter 120 can be a user terminal device and the receiver 160 can be a radio unit, but the invention is not limited thereto. For example, the transmitter 120 can be a user terminal and the receiver 160 can be a base station.

[0081] Figure 8 This is a schematic diagram of a communication method 80 according to an embodiment of the present invention. The communication method 80 can be used at the receiving end 160. The communication method 80 can be compiled into program code and executed by a processing circuit, and stored in a storage circuit. The communication method 80 may include the following steps:

[0082] Step S800: Start.

[0083] Step S802: Receive a received signal (e.g., received signal Y or Y0) * ), wherein the transmitting end 120 is based on a raw signal (e.g., raw signal Z or Z0). * A transmitted signal (e.g., transmitted signal X or X') is obtained. * The transmitting end 120 sends a transmission signal, which is then converted into a receiving signal through channel 140.

[0084] Step S804: Obtain information about the original signal based on the received signal, wherein a structural causal model (e.g., causal model 20) is used to associate the transmitted signal and the received signal, and multiple causal variables of a causal graph (e.g., causal graph CG) of the structural causal model between the transmitted signal and the received signal and a causal structure of the causal graph are determined together.

[0085] Step S806: End.

[0086] Figure 9 This is a schematic diagram of a communication method 90 according to an embodiment of the present invention. The communication method 90 can be used at the transmitting end 120. The communication method 90 can be compiled into program code and executed by a processing circuit, and stored in a storage circuit. The communication method 90 may include the following steps:

[0087] Step S900: Begin.

[0088] Step S902: Based on an original signal (e.g., original signal Z or Z0), * A transmitted signal (e.g., transmitted signal X or X') is obtained. * ).

[0089] Step S904: Send a transmission signal, wherein the transmission signal is transformed into a received signal (e.g., received signal Y or Y) through channel 140. *The receiver 160 obtains information about the original signal based on the received signal and uses a structural causal model (e.g., causal model 20) to associate the transmitted signal with the received signal. Multiple causal variables of a causal graph (e.g., causal graph CG) of the structural causal model between the transmitted signal and the received signal, as well as a causal structure of the causal graph, are determined together.

[0090] Step S906: End.

[0091] In one embodiment, the storage circuitry can be used to store image data or instructions. The storage circuitry may be a Subscriber Identity Module (SIM), Read-Only Memory (ROM), Flash memory, Random-Access Memory (RAM), CD-ROM / DVD-ROM / BD-ROM, Magnetic tape, Hard disk, Optical data storage device, Non-volatile storage device, Non-transitory computer-readable medium, and is not limited thereto.

[0092] In one embodiment, the processing circuitry can be used to execute instructions, which may be a central processing unit, a microprocessor, or an application-specific integrated circuit (ASIC), but is not limited thereto.

[0093] In summary, this invention utilizes causal relationships to obtain the original signal, rather than relying on statistical correlation. This allows for accurate correlation between the received signal and the transmitted signal, thereby reducing the transmission error rate. Furthermore, this invention eliminates the need for channel estimation at the user end to obtain the original signal, thus reducing power consumption and processor utilization.

[0094] The above description is only a preferred embodiment of the present invention. All equivalent changes and modifications made in accordance with the claims of the present invention should be included within the scope of the present invention.

Claims

1. A communication method for a receiving end, comprising: receiving a received signal, wherein A transmitting end obtains a transmitted signal based on an original signal, transmits the transmitted signal, and the transmitted signal is transformed into the received signal through a channel, wherein the original signal and the transmitted signal are substantially the same; and Information about the original signal is obtained based on the received signal. A structural causal model is used to link the transmitted signal and the received signal. Multiple causal variables and the causal structure of a causal graph of this structural causal model are determined together between the transmitted signal and the received signal. The received signal is back-reasoned into multiple combinations based at least on abductive reasoning. Each of the multiple combinations includes a candidate transmission signal and a noise value. A preferred transmission signal is selected from the multiple candidate transmission signals to correspond to the transmission signal based on a signal-to-noise ratio, thereby converting the received signal into the original signal based at least on abductive reasoning.

2. The communication method as described in claim 1, wherein, This causal graph is generated based on maximum a posteriori and point estimation.

3. The communication method as described in claim 1, wherein, By using multiple observation functions to map multiple data points in a base dataset to the multiple causal variables in the causal graph, the causal graph is generated from the base dataset based on maximum a posteriori and point estimation.

4. The communication method as described in claim 3, wherein, These multiple observation functions are obtained based on a causal semantic generation model.

5. The communication method as described in claim 1, wherein, The receiving end is either a user terminal or a radio unit, and the transmitting end is either the user terminal or the radio unit.

6. The communication method as described in claim 1, wherein, The received signal includes the distorted transmitted signal and noise, and at least one noise value corresponding to the structural causal model is set as the mean of the noise.

7. A communication device for a receiving end, comprising: A storage circuit for storing an instruction, which includes: Receive a received signal, wherein, A transmitting end obtains a transmitted signal based on an original signal, transmits the transmitted signal, and the transmitted signal is transformed into the received signal through a channel, wherein the original signal and the transmitted signal are substantially the same; and Information about the original signal is obtained based on the received signal, wherein a structural causal model is used to associate the transmitted signal and the received signal. Multiple causal variables of a causal graph in the structural causal model and a causal structure of the causal graph are determined together. The received signal is back-reasoned into multiple combinations, at least based on abductive reasoning. Each of these combinations includes a candidate transmitted signal and a noise value. A preferred transmitted signal is selected from the multiple candidate transmitted signals based on a signal-to-noise ratio, thereby converting the received signal into the original signal at least based on abductive reasoning. A processing circuit, coupled to the storage circuit, is used to execute the instruction stored in the storage circuit.

8. A communication method for a transmitting end, comprising: A transmitted signal is obtained from an original signal, and the original signal is substantially the same as the transmitted signal. Send the transmission signal, wherein, The transmitted signal is transformed into a received signal through a channel. A receiving end obtains information about the original signal based on the received signal. A structural causal model is used to associate the transmitted signal and the received signal. The multiple causal variables of a causal graph in the structural causal model and the causal structure of the causal graph are determined together. The received signal is back-reasoned into multiple combinations based at least on abductive reasoning. Each of the multiple combinations includes a candidate transmission signal and a noise value. A preferred transmission signal is selected from the multiple candidate transmission signals to correspond to the transmission signal based on a signal-to-noise ratio, thereby converting the received signal into the original signal based at least on abductive reasoning.

9. The communication method as described in claim 8, wherein, This causal graph is generated based on maximum a posteriori and point estimation.

10. The communication method as described in claim 8, wherein, By using multiple observation functions to map multiple data points in a base dataset to the multiple causal variables in the causal graph, the causal graph is generated from the base dataset based on maximum a posteriori and point estimation.

11. The communication method as described in claim 10, wherein, These multiple observation functions are obtained based on a causal semantic generation model.

12. The communication method as described in claim 8, wherein, The receiving end is either a user terminal or a radio unit, and the transmitting end is either the user terminal or the radio unit.

13. The communication method as described in claim 8, wherein, The received signal includes the distorted transmitted signal and noise, and at least one noise value corresponding to the structural causal model is set as the mean of the noise.

14. A communication device for a transmitting end, comprising: A storage circuit for storing an instruction, which includes: A transmitted signal is obtained from an original signal, and the original signal is substantially the same as the transmitted signal. Send the transmission signal, wherein, The transmitted signal is transformed into a received signal through a channel. A receiving end obtains information about the original signal based on the received signal and uses a structural causal model to associate the transmitted signal and the received signal. Multiple causal variables and a causal structure of a causal graph in the structural causal model are determined together. The received signal is back-reasoned into multiple combinations, at least based on abductive reasoning. Each of these combinations includes a candidate transmitted signal and a noise value. A preferred transmitted signal is selected from the multiple candidate transmitted signals based on a signal-to-noise ratio, thereby converting the received signal back into the original signal, at least based on abductive reasoning. A processing circuit, coupled to the storage circuit, is used to execute the instruction stored in the storage circuit.

Images