Low sampling bit based anti-multipath IR-UWB frequency offset estimation method

Abstract

The application provides a low-sampling-bit-based anti-multipath IR-UWB frequency offset estimation method and relates to the technical field of IR-UWB. The application comprises the following steps: adopting a constant false alarm rate method to sample a received signal; intercepting a first symbol period part of a preamble symbol sequence of the received signal as a matched filter template and performing matched filtering; extracting a correlation peak value of an output result of the matched filter; normalizing the correlation peak value to reserve amplitude angle information, obtaining a phase value, and demisting to obtain a frequency offset sequence; adopting Bayesian estimation to perform iteration and correction to obtain a final frequency offset estimation value; and correcting a local oscillator of a receiving end by generating a sine wave and a cosine wave corresponding to the estimated frequency offset. The application saves hardware resources and improves algorithm real-time performance by reducing the sampling bit number; improves anti-multipath performance in a complex multipath environment by using a real-time collected matched filter template; and improves frequency offset estimation precision by adopting Bayesian estimation.

Description

Technical Field

[0001] This invention relates to the field of IR-UWB technology, and in particular to an IR-UWB frequency offset estimation method at low sampling bit depth. Background Technology

[0002] Impulse Radio-UltraWideband (IR-UWB) technology uses extremely narrow pulse signals at the nanosecond level to transmit data, offering high time resolution and excellent multipath resistance, enabling precise ranging and positioning. However, clock frequency offset exists between terminals and base stations with independent clocks, which is a significant factor affecting the positioning accuracy of IR-UWB, thus requiring estimation and correction.

[0003] One common method for frequency offset estimation in current IR-UWB systems is to extract the timestamps of two ranging measurements and use the ratio of the arrival time difference to the time interval between the transmitted signal as the estimated frequency offset. However, noise and multipath interference can severely affect the estimation of the arrival time. Another method is to use a local matching template for correlation processing to obtain the phase difference. However, local templates are not suitable for complex multipath environments, and the high sampling bit depth consumes a lot of hardware resources, which cannot meet the requirements of high real-time systems.

[0004] Therefore, it is necessary to propose a low-complexity frequency offset estimation method that is resistant to multipath propagation to address the problems existing in the above methods. Summary of the Invention

[0005] This invention proposes a multipath-resistant IR-UWB frequency offset estimation method based on low sampling bit depth. By reducing the sampling bit depth, hardware resources are saved and the real-time performance of the algorithm is improved. The real-time acquired matched filter template is used to improve the multipath resistance performance in complex multipath environments. Bayesian estimation is used to improve the frequency offset estimation accuracy.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0007] A method for estimating frequency offset against multipath IR-UWB based on low sampling bit depth includes the following steps:

[0008] (1) Mix the received signal with the local oscillator of the receiving end, perform down-conversion and low-pass filtering, and use the constant false alarm probability method to select a threshold to sample the received signal in 1.5 bits;

[0009] (2) The first symbol period of the received signal preamble sequence is extracted as the template for matched filtering and matched filtering is performed. The output of the matched filtering will show periodic correlation peaks with phase shift.

[0010] (3) Extract the relevant peaks from the matched filter output. The number of peaks is equal to the number of leading symbols, resulting in the relevant peak sequence [corr(0), corr(1), corr(2), ..., corr(num)]. sync -1)];

[0011] Where, num sync The number of symbols for the SYNC preamble;

[0012] (4) Normalize the relevant peak values ​​and retain the argument information, calculate the phase value and deblur the information to obtain the phase sequence, and use the following formula to obtain the frequency offset sequence.

[0013]

[0014] Where f is the true value of the frequency offset, and d∈{0,1,…,num} sync}, T s The sampling interval is... This refers to the frequency offset error;

[0015] (5) The frequency offset sequence is iterated and corrected using Bayesian estimation to obtain the frequency offset estimate:

[0016]

[0017] Where arg max represents the frequency offset value obtained when the probability is maximized. For conditional probability;

[0018] (6) Generate sine and cosine waves corresponding to the estimated frequency offset to correct the local oscillator at the receiving end.

[0019] Furthermore, a constant false alarm rate (CFAR) method is used to select a threshold for 1.5-bit sampling of the received signal. After sampling, the signal contains only three symbols: {-1, 0, 1}. The discrete signal after sampling is represented as follows:

[0020]

[0021] Among them, B ik ∈{0, ±1}, w(n-τ) l ( ) represents the received signal waveform. The signal is the multipath channel impulse response, t0 is the signal arrival time, and α is the signal arrival time. l τ l Here, L and l represent the multipath fading coefficient and multipath delay, respectively, and the multipath number and multipath number number index, respectively. Let T be the initial phase of the carrier, Δf be the frequency offset, and T be the initial phase of the carrier. s T is the sampling interval time. p Let ξ(n) be the chip time interval, ξ(n) be additive white Gaussian noise, and ck For PN code sequence numbering, c k ∈{0, 1, ..., K}, where K is the number of leading symbols.

[0022] Furthermore, the first symbol period of the received signal preamble sequence is extracted as the matched filter template, specifically as follows:

[0023] (1) Set a threshold to determine the arrival time of the first preamble symbol;

[0024] (2) Using 1.5-bit sampling, the sampled signal of the first symbol period is extracted at the time of signal arrival; the extraction length is greater than the symbol period.

[0025] (3) Repeat the above operation after the next ranging signal arrives to generate a 1.5-bit low-bit real-time dynamic matching template.

[0026] Furthermore, Bayesian estimation is used to iterate and correct the frequency offset sequence to obtain the final frequency offset estimate. Specifically:

[0027] Let frequency offset f be a random variable, and the frequency offset estimation sequence be... in, Let d be the phase noise variance, and d be the leading symbol number. The frequency offset is used to estimate the variance in the sequence. The probability distribution function is used as initial prior information:

[0028]

[0029] at this time, The first maximum a posteriori probability estimate is performed using the observations; the conjugate prior distribution of the normal mean is a normal distribution, thus yielding the posterior distribution:

[0030]

[0031] Its expected value f2 and variance They are respectively:

[0032]

[0033]

[0034] Thus, the frequency offset estimate is obtained.

[0035]

[0036] Then, As the prior distribution for the next iteration, Using the observations as the basis for maximum a posteriori probability estimation, the posterior distribution is obtained. Iterate through numsync -2 times, to obtain the final posterior distribution:

[0037]

[0038] Its expectation f d and variance They are respectively:

[0039]

[0040]

[0041] Under the maximum a posteriori probability, the frequency offset estimate is calculated.

[0042]

[0043] in, This represents the conditional probability.

[0044] Compared with the prior art, the present invention has the following advantages:

[0045] (1) It has good anti-multipath capability. The present invention extracts the first period of the preamble symbol of the received signal as the matched filter template, which has higher coherence gain in complex multipath environments.

[0046] (2) Ensure real-time positioning and communication. The algorithm has low complexity and consumes few hardware resources. This invention uses a 1.5-bit sampling ADC, which greatly improves the calculation speed in the matched filtering process. In the frequency offset estimation process, only the coherent peak value corresponding to the preamble symbol is extracted, which fully saves storage resources.

[0047] (3) Bayesian estimation improves estimation accuracy. While extracting the coherence peak of a single symbol, the Bayesian estimation algorithm is used to iteratively correct the prior information in real time. During the iteration process, the frequency offset estimate tends to be stable, which significantly improves the estimation accuracy under 1.5-bit sampling. Attached Figure Description

[0048] Figure 1 This is a flowchart of the matched filtering process based on 1.5-bit sampling in an embodiment of the present invention.

[0049] Figure 2 This is a flowchart of frequency offset estimation in an embodiment of the present invention.

[0050] Figure 3 This is a flowchart of the Bayesian frequency offset estimation in an embodiment of the present invention. Detailed Implementation

[0051] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0052] A method for frequency offset estimation in multipath IR-UWB based on low sampling bit depth. Figure 1 This is a flowchart of a matched filter based on 1.5-bit sampling, including the following steps:

[0053] (1) The received signal is down-converted and low-pass filtered. The constant false alarm probability method is used to select an appropriate threshold to sample the I and Q signals 1.5 bits respectively.

[0054] This sampling method uses a dynamic threshold, which adjusts the sampling threshold in real time according to the noise level.

[0055] (2) Extract the first symbol period of the received signal preamble as a matched filter template and perform matched filtering. After detecting the first path of the first preamble symbol, store the template signal with a storage length equal to the period of a single preamble symbol.

[0056] The matched filter template is represented as:

[0057]

[0058] The sampling frequency f in this embodiment s =1.9936GHz. According to the UWB protocol, the length of the spreading code within a symbol is 31 or 127. Choosing 127 can obtain a better correlation peak.

[0059] (3) Extract the correlation peaks from the matched filter output. The number of peaks is equal to the number of leading symbols. The resulting correlation peak sequence is represented as follows:

[0060] [corr(0), corr(1), corr(2),…, corr(num sync -1)]

[0061] Where, num sync The number of preamble symbols is 64, according to the UWB protocol, in order to make the frequency offset estimation more accurate.

[0062] Figure 2 This is a flowchart of the frequency offset estimation process in an embodiment of the present invention, and its steps are as follows:

[0063] (1) The extracted relevant peak sequences are normalized using the following formula:

[0064]

[0065] Where d is the relevant peak sequence number, and n r This represents the frequency offset estimation error. At this point, the relevant peak sequence retains only the phase information and the error term.

[0066] (2) After normalizing the relevant peak values, only the argument information is retained. The phase value is then calculated and defuzzified to obtain the phase sequence.

[0067] The received signal of the first symbol period is selected as the matching template, so the first phase value to be solved is 0.

[0068] (3) The frequency offset sequence is obtained using the following formula.

[0069]

[0070] Where f is the true value of the frequency offset, and d∈{0,1,…,num} sync}, This represents the frequency offset error.

[0071] (4) The frequency offset sequence is iterated and corrected using the Bayesian estimation algorithm to obtain the final frequency offset estimate:

[0072]

[0073] in, This represents the conditional probability.

[0074] (5) Generate sine and cosine waves corresponding to the estimated frequency offset to correct the local oscillator at the receiving end.

[0075] The correction process is calculated using the following formula:

[0076]

[0077] After selecting the matching template in this embodiment, its initial phase Since the initial phase is zero, the step of estimating the initial phase can be omitted.

[0078] Figure 3 This is a flowchart of the Bayesian frequency offset estimation algorithm in an embodiment of the present invention. It details the frequency offset estimation process using Bayesian maximum a posteriori probability after obtaining the frequency offset sequence.

[0079] After initializing the prior information, the Bayesian estimation process is initiated sequentially based on the frequency offset values ​​obtained from the time series. Detailed explanation follows:

[0080] Let frequency offset f be a random variable, and the frequency offset estimation sequence be... in, Let d be the phase noise variance, and d be the leading symbol number. The frequency offset is used to estimate the variance in the sequence. The probability distribution function is used as initial prior information:

[0081]

[0082] at this time, The first maximum a posteriori probability estimate is performed using the observations. The conjugate prior distribution of the normal mean (with known variance) is a normal distribution; therefore, the posterior distribution is obtained:

[0083]

[0084] Its expected value f2 and variance They are respectively:

[0085]

[0086]

[0087] Therefore, the frequency offset estimate is obtained.

[0088]

[0089] Then, As the prior distribution for the next iteration, Using the observations as the basis for maximum a posteriori probability estimation, the posterior distribution is obtained. Iterate through num in this way sync -2 times, to obtain the final posterior distribution:

[0090]

[0091] Its expectation f d and variance They are respectively:

[0092]

[0093]

[0094] Under the maximum a posteriori probability, the frequency offset estimate is calculated.

[0095]

[0096] in, This represents the conditional probability.

[0097] This invention employs 1.5-bit low-bit sampling, extracts the received signal of the first preamble symbol as a matching template, obtains the normalized correlation peak sequence, calculates the frequency offset sequence, and then enters the Bayesian estimation process in time sequence to finally obtain the optimal frequency offset estimate under the maximum a posteriori probability, thus achieving high-precision frequency offset estimation and correction.

[0098] In summary, this invention saves hardware resources and improves the real-time performance of the algorithm by reducing the number of sampling bits; it improves the anti-multipath performance in complex multipath environments by using a real-time acquired matched filter template instead of a local template; and it improves the frequency offset estimation accuracy by using Bayesian estimation.

Claims

1. A method for estimating frequency offset against multipath IR-UWB based on low sampling bit depth, characterized in that, Includes the following steps: (1) Mix the received signal with the local oscillator of the receiving end, perform down-conversion and low-pass filtering, and use the constant false alarm probability method to select the threshold to sample the received signal in 1.5 bits; (2) The first symbol period of the received signal preamble sequence is extracted as the template for matched filtering and matched filtering is performed. The output of the matched filtering will show periodic correlation peaks with phase shift. (3) Extract the relevant peaks from the matched filter output. The number of peaks is equal to the number of leading symbols, resulting in the relevant peak sequence [corr(0), corr(1), corr(2), ..., corr(num)]. sync -1)]; Where, num sync The number of symbols for the SYNC preamble; (4) Normalize the relevant peak values ​​and retain the argument information, calculate the phase value and deblur the information to obtain the phase sequence, and use the following formula to obtain the frequency offset sequence. Where f is the true value of the frequency offset, and d∈{0,1,…,num} sync }, T s The sampling interval is... This refers to the frequency offset error; (5) The frequency offset sequence is iterated and corrected using Bayesian estimation to obtain the frequency offset estimate: Where arg max represents the frequency offset value obtained when the probability is maximized. For conditional probability; (6) Generate sine and cosine waves corresponding to the estimated frequency offset to correct the local oscillator at the receiving end.

2. The method for frequency offset estimation of multipath IR-UWB based on low sampling bit depth as described in claim 1, characterized in that, A constant false alarm rate (CFAR) method is used to select a threshold for 1.5-bit sampling of the received signal. After sampling, the signal contains only three symbols: {-1, 0, 1}. The discrete signal after sampling is represented as: Among them, B ik ∈{0, ±1}, w(n-τ) l ( ) represents the received signal waveform. The signal is the multipath channel impulse response, t0 is the signal arrival time, and α is the signal arrival time. l τ l Here, L and l represent the multipath fading coefficient and multipath delay, respectively, and the multipath number and multipath number number index, respectively. Let T be the initial phase of the carrier, Δf be the frequency offset, and T be the initial phase of the carrier. s T is the sampling interval time. p Let ξ(n) be the chip time interval, ξ(n) be additive white Gaussian noise, and c k For PN code sequence numbering, c k ∈{0, 1, ..., K}, where K is the number of leading symbols.

3. The method for frequency offset estimation of multipath IR-UWB based on low sampling bit depth according to claim 1, characterized in that, The first symbol period of the received signal preamble is extracted as the matched filter template, specifically as follows: (1) Set a threshold to determine the arrival time of the first preamble symbol; (2) Using 1.5-bit sampling, the sampled signal of the first symbol period is extracted at the time of signal arrival; the extraction length is greater than the symbol period. (3) Repeat the above operation after the next ranging signal arrives to generate a 1.5-bit low-bit real-time dynamic matching template.

4. The method for frequency offset estimation of multipath IR-UWB based on low sampling bit depth according to claim 1, characterized in that, Bayesian estimation is used to iterate and correct the frequency offset sequence to obtain the final frequency offset estimate. The specific method is as follows: Let the frequency offset f be a random variable, and the frequency offset estimation sequence be... in, Let d be the phase noise variance, and d be the leading symbol number. The frequency offset is used to estimate the variance in the sequence. The probability distribution function is used as initial prior information: at this time, The first maximum a posteriori probability estimate is performed using the observations; the conjugate prior distribution of the normal mean is a normal distribution, thus yielding the posterior distribution: Its expected value f2 and variance They are respectively: Thus, the frequency offset estimate is obtained. Then, As the prior distribution for the next iteration, Using the observations as the basis for maximum a posteriori probability estimation, the posterior distribution is obtained. Iterate through num sync -2 times, to obtain the final posterior distribution: Its expected f d and variance They are respectively: Under the maximum a posteriori probability, the frequency offset estimate is calculated. in, This represents the conditional probability.

Images