A time-varying underwater acoustic channel estimation method based on Kalman filter

Abstract

The application relates to a time-varying underwater acoustic channel estimation method based on Kalman filtering. The application belongs to the field of underwater acoustic channel estimation, and specifically, the time correlation of the underwater acoustic channel and the cluster sparsity of the underwater acoustic channel are utilized to improve a traditional KF-CS algorithm, the selection process of a random change support set of the traditional KF-CS algorithm is removed, the running time of the algorithm can be greatly simplified, and meanwhile, the correlation between channels can be effectively utilized to improve the channel estimation performance. The specific implementation flowchart of the application is shown in FIG. 1 in the description.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of underwater acoustic channel estimation, and specifically utilizes the sparse cluster characteristics of underwater acoustic channel and the time correlation of underwater acoustic channel to improve the traditional KF-CS algorithm, removes the selection process of the traditional KF-CS algorithm for the randomly changing support set, greatly simplifies the running time of the algorithm, and effectively utilizes the correlation between channels to improve the performance of channel estimation. BACKGROUND

[0002] Underwater acoustic communication is currently the only way to realize underwater long-distance and high-data communication, but the underwater acoustic channel is a time-varying channel, and the available frequency spectrum resources are limited, which seriously limits the development of underwater high-data communication. With the rapid development of technology, the orthogonal frequency division multiplexing technology is widely used in land 4G communication due to its high spectrum utilization and anti-multipath effect. At present, a large number of scholars are actively studying how to apply OFDM technology to underwater acoustic communication, trying to break through the limitations of underwater environment to realize high-speed data communication.

[0003] In order to realize reliable OFDM underwater communication, accurate channel estimation technology is necessary. Studies have shown that the underwater acoustic channel is sparse, and the compressive sensing model is mainly aimed at the recovery of sparse signals, so most of the current researches consider using compressive sensing algorithm to reconstruct the underwater acoustic channel. The classic compressive sensing recovery algorithms include orthogonal matching pursuit algorithm, sparse adaptive matching pursuit algorithm, and basis pursuit algorithm based on convex optimization principle, and these algorithms have generated a large number of improved algorithms when combined with underwater acoustic channel estimation.

[0004] Since the self-distributed compressive sensing framework was proposed, a large number of scholars have begun to focus on the time-space correlation between time-varying underwater acoustic channels, and have proposed corresponding improved algorithms to improve the performance of channel estimation. With the deepening of people's understanding of underwater acoustic channels, studies have shown that the underwater acoustic channel does not present a common sparse form, but a cluster sparse model, that is, the non-zero coefficients in the underwater acoustic channel are non-uniformly gathered in some areas.

[0005] Kalman filtering and compressive sensing have been studied by scholars since they were proposed. Therefore, some scholars have proposed a Kalman filtering compressive sensing algorithm, which uses Kalman filtering to recover the compressive sensing signal. However, this method is relatively complex to implement, mainly because the position of the current channel sparse tap cannot be directly determined, and the changing support set must be determined by calculating the residual. Some scholars have also proved the feasibility of using KF-CS algorithm to recover underwater acoustic channel, while using the original dual pursuit algorithm to determine the changing support set, but there is still a certain degree of complexity.

[0006] The application mainly aims at the simplification process of KF-CS algorithm, and mainly determines the position of the stable support set in the underwater acoustic channel through the channel information estimated at the previous moment and the Kalman prediction at the current moment, simultaneously determines the position of the candidate support set of the current channel by using the cluster sparse characteristics of the underwater acoustic channel, and finally calculates the sparse size on the candidate support set of the channel by using the Kalman filtering method, and filters out the wrong atoms by setting the threshold value, so as to achieve accurate channel estimation. SUMMARY

[0007] The application mainly aims at the channel estimation technology of underwater acoustic communication, and by improving the traditional KF-CS algorithm, the complexity of the traditional KF-CS algorithm is simplified under the premise of ensuring the channel estimation performance, the running time of the algorithm is greatly reduced, and the real-time performance of the algorithm is enhanced.

[0008] The application adopts the following scheme:

[0009] Step 1: input the received OFDM pilot signal y, input the channel impulse response at the previous moment Input the observation matrix D p , input the covariance matrix P of the channel at the previous moment t-1 and the channel noise variance

[0010] Step 2: according to the input at the previous moment, judge the corresponding candidate set and length Set the covariance matrix of the observation noise Suppose that the covariance matrix of the process noise is According to the size of , set the measurement matrix Set the error threshold

[0011] Step 3: KF prediction, calculate the gain coefficient K of the Kalman filter at the current moment according to formula (1) t,tmp Then calculate the predicted value of the channel at the current moment by formula (2)

[0012]

[0013] Step 4: filter out the multipath whose coefficient in is less than , and the remaining paths are considered to have stable time delay, so as to obtain the candidate set and length

[0014] Step 5: KF update, initialize determine and The prediction error covariance matrix at time t is calculated using equation (3), the gain coefficient of the Kalman filter is updated using equation (4), and finally the channel covariance matrix and channel information at the current time are updated using equations (5) and (6).

[0015] P t|t-1 =P t-1 +Q t (3)

[0016]

[0017] P t =(IK t α t )P t|t-1 (5)

[0018]

[0019] Step 6: Filter out error atoms, if The minimum value in is less than Then proceed to the error filtering procedure to remove... Less than Part of the update will be made simultaneously. Q t and α t Then proceed to the third step until... The median is greater than Then stop iterating;

[0020] The advantages of this invention are:

[0021] 1. The performance of the algorithm is improved by utilizing the cluster sparsity characteristic of the underwater acoustic channel.

[0022] 2. It can effectively utilize the time correlation between underwater acoustic channels to improve the algorithm's performance.

[0023] 3. The operation steps of the traditional KF-CS algorithm are simplified, and the running time of the algorithm is greatly reduced while ensuring the performance remains unchanged. Attached Figure Description

[0024] Figure 1 This is a flowchart of the present invention. Detailed implementation method:

[0025] The invention will now be described in detail with reference to specific examples.

[0026] Step 1: Input the received OFDM pilot signal y, where the length of the pilot signal is 100, and input the channel impulse response at the previous time step. Input observation matrix Dp Input the channel covariance matrix P from the previous time step. t-1 and channel noise variance

[0027] Step 2: Based on the input from the previous moment To determine the corresponding candidate set and length Set the covariance matrix of the observation noise. Assume the covariance matrix of the process noise is in according to Size setting measurement matrix Set error threshold

[0028] Step 3: Perform KF prediction and calculate the gain coefficient K of the Kalman filter at the current time according to equation (1). t,tmp Then, the predicted value of the channel at the current time is calculated using equation (2).

[0029]

[0030] Step 4: Filtering The coefficient is less than In a multipath scenario, all remaining paths are considered to have stable time delays, thus obtaining the candidate set at time t. and length

[0031] Step 5: KF update, initialization Sure and The prediction error covariance matrix at time t is calculated using equation (3), the gain coefficient of the Kalman filter is updated using equation (4), and finally the channel covariance matrix and channel information at the current time are updated using equations (5) and (6).

[0032] P t|t-1 =P t-1 +Q t (3)

[0033]

[0034] P t =(IK t α t )P t|t-1 (5)

[0035]

[0036] Step 6: Filter out error atoms, if The minimum value in is less than Then proceed to the error filtering procedure to remove... Less than Part of the update will be made simultaneously. Q t and α t Then proceed to step three again, until h t The median is greater than Then stop iterating.

Claims

1. A time-varying underwater acoustic channel estimation method based on Kalman filtering, characterized in that... The following steps are required: Step 1: Input the received OFDM pilot signal y, and input the channel impulse response from the previous moment. Input observation matrix D p Input the channel covariance matrix P from the previous time step. t-1 and channel noise variance Step 2: Based on the input from the previous moment To determine the corresponding candidate set and length Set the covariance matrix of the observation noise. Assume the covariance matrix of the process noise is according to Size setting measurement matrix Set error threshold Step 3: Perform KF prediction and calculate the gain coefficient K of the Kalman filter at the current time according to equation (1). t,tmp Then, the predicted value of the channel at the current time is calculated using equation (2). Step 4: Filtering The coefficient is less than In a multipath scenario, all remaining paths are considered to have stable time delays, thus obtaining the candidate set at time t. and length Step 5: KF update, initialization Sure and The prediction error covariance matrix at time t is calculated using equation (3), the gain coefficient of the Kalman filter is updated using equation (4), and finally the channel covariance matrix and channel information at the current time are updated using equations (5) and (6). P t|t-1 =P t-1 +Q t (3) P t =(I-K t α t )P t|t-1 (5) Step 6: Filter out error atoms, if The minimum value in is less than Then proceed to the error filtering procedure to remove... Less than Part of the update will be made simultaneously. Q t and α t Then proceed to the third step until... The median is greater than Then stop iterating.

Images