阅读说明：本技术 一种基于星载ais系统的信号检测与频偏估计算法 (Signal detection and frequency offset estimation algorithm based on satellite-borne AIS system ) 是由 戴雨峰 于 2021-01-08 设计创作，主要内容包括：本发明公开了一种基于星载AIS系统的信号检测与频偏估计算法,在相关检测前,先将差分后的接收序列需要减去自身均值,即减去直流分量,再与本地序列进行相关,则在频偏指标要求范围内,其相关峰值和峰值特征均与频偏大小几乎无关,可在不同频偏条件下使用统一的检测器结构。本发明将自适应门限相关检测法与峰值特征匹配法相结合,次峰由于特征不匹配被筛除。两者结合克服了由于训练序列为非伪随机码带来的次峰干扰问题,准确检测到最佳峰值位置。本发明可以有效对抗星载条件下的多普勒频偏大的问题,在指标要求的-4kHz～+4kHz范围内,漏警概率远低于系统丢包率,且系统丢包率跟频偏大小无明显关系。(The invention discloses a signal detection and frequency offset estimation algorithm based on a satellite-borne AIS system, wherein before relevant detection, the mean value of a receiving sequence after difference is required to be subtracted, namely, the direct current component is subtracted, and then the receiving sequence is correlated with a local sequence, so that the relevant peak value and the peak value characteristics are almost irrelevant to the frequency offset within the range required by a frequency offset index, and a unified detector structure can be used under different frequency offset conditions. The invention combines the self-adaptive threshold correlation detection method with the peak value characteristic matching method, and secondary peaks are screened out due to characteristic mismatching. The two are combined to overcome the problem of secondary peak interference caused by non-pseudo random codes of the training sequence, and the optimal peak position is accurately detected. The invention can effectively solve the problem of large Doppler frequency offset under the satellite-borne condition, the false alarm missing probability is far lower than the system packet loss rate within the range of-4 kHz to +4kHz required by indexes, and the system packet loss rate has no obvious relation with the frequency offset.)

1. A signal detection and frequency offset estimation algorithm based on a satellite-borne AIS system is characterized by comprising the following steps:

step one, converting an analog input signal into a digital signal after AD sampling, passing through a data delayer, directly performing quadrature frequency mixing after AD sampling if no signal is detected, and otherwise performing quadrature frequency mixing on the signal after data delay;

secondly, the mixed signal passes through a low-pass filter to filter out-of-band noise and image frequency spectrum, and the bandwidth of the filter contains the signal at the time of maximum frequency offset;

step three, carrying out data differential operation, dividing the differential signals into three paths, and respectively sending the three paths of signals to a relevant detector, a frequency offset estimation module and a matched filtering module;

step four, when the output of the correlation detector meets the peak value and is larger than the adaptive threshold, the characteristic matching is carried out, and only when the characteristic error is smaller than the set threshold, the signal is detected;

step five, if the signal is detected, finding a frame header, and performing three operations, namely performing frequency offset estimation by using the cached differential data and adjusting DDS frequency control words to perform frequency compensation according to an estimation result; secondly, performing initial phase compensation; thirdly, re-entering the training sequence and the frame header which have finished the frequency offset compensation, so that the frequency is compensated and the bit synchronization is locked in the time before the next frame header arrives;

and step six, performing matched filtering, bit synchronization, data judgment, frame head and frame tail detection and CRC (cyclic redundancy check) operation after difference.

2. The signal detection and frequency offset estimation algorithm based on the spaceborne AIS system as claimed in claim 1, wherein in the third step, before the correlation detection, the differentiated received sequence needs to subtract its mean value, i.e. subtract the direct current component, and then is correlated with the local sequence.

3. The signal detection and frequency offset estimation algorithm based on the spaceborne AIS system as claimed in claim 1, wherein in the fourth step, the adaptive threshold is obtained by adopting a radar field constant false alarm rate detection structure.

4. A method as claimed in claim 1, based onAnd in the fourth step, assuming that L secondary peaks in front of the main peak are considered, and the ratios of the secondary peaks to the main peak are respectively lambda_i0Wherein i is 1, 2,. L; when the peak value is detected and is larger than the adaptive threshold, calculating the ratio lambda of the secondary peak value and the primary peak value of the fixed position before the peak value_iThen, the sum of the squared errors is calculated according to the following formula:

if e is less than delta and delta is a set threshold, the characteristic is considered to be met, and a signal is detected;

if e > δ, the signature is deemed not to be satisfied, indicating that no signal has been detected.

5. The signal detection and frequency offset estimation algorithm based on the spaceborne AIS system as claimed in claim 1, wherein in the fifth step, the initial phase compensation comprises adjusting the phase of the bit synchronization loop.

6. The signal detection and frequency offset estimation algorithm based on the satellite-borne AIS system as claimed in claim 1, wherein in the fifth step, the training sequence and the frame header with completed frequency offset compensation are re-entered, i.e. the data delay switch is switched, so that the data input to the quadrature mixer is the data after delaying several symbols, and the frequency is compensated and the bit synchronization is locked in the time before the next frame header arrives.

Technical Field

The invention relates to the field of ship communication, in particular to a signal detection and frequency offset estimation algorithm based on a satellite-borne AIS system.

Background

With the continuous development of shipping services, the existing maritime communication technology is difficult to meet the increasing user requirements, and the VDES-based "aerospace-marine" communication network becomes a hotspot of international maritime communication research. The 'aerospace, ground and sea' network aims to perfect communication among ships and banks, ships and satellites in marine communication and make up for the shortages of coverage, communication quality and the like of marine communication in China. The research object of the invention is an AIS subsystem of space-based VDES in an 'aerospace-geodetic' network, in particular to a satellite-borne AIS receiver for communication between ships and satellites in marine communication.

The development of the satellite-borne AIS receiver needs to solve two key problems:

firstly, aiming at the burst characteristic of the AIS signal, it is necessary to design a reliable signal detection algorithm based on a training sequence, so that after the signal is detected, parameters such as frequency, phase and the like can be estimated and compensated according to buffered data, and operations such as data buffering, time delay and the like are combined to ensure that the frame header and the data have no frequency offset and are locked in a bit synchronization manner before arriving.

Secondly, compared with the conventional ship-shore and ship-to-ship communication, the communication Doppler frequency offset between the ship and the satellite is larger and reaches-4 kHz to +4kHz, and the maximum frequency offset is about 42 percent of the data rate, so that for a satellite-borne receiver, the estimation and compensation of the larger Doppler frequency offset are also one of the key problems which need to be solved.

The method solves the problem of unstable false alarm (or false missing) probability caused by overlarge dynamic range of the signal and improves the stability of signal detection. However, in a satellite-borne AIS application scenario, the traditional correlation detection method has two disadvantages:

on one hand, the method generally requires that the training sequence satisfies the pseudo-random characteristic, while the training sequence of the AIS signal is a 01-alternating sequence, which is not a pseudo-random code, and a plurality of secondary peaks are generated near the maximum peak value during the correlation detection, thereby reducing the resolution capability of the optimal peak value.

On the other hand, when the received signal is directly subjected to correlation detection, the correlation peak value is reduced along with the increase of the frequency offset, and even no correlation peak value is output. Although the traditional method adopting a plurality of parallel correlators can solve the problem theoretically, the traditional method has high computational complexity and large resource overhead, and is not recommended to be adopted in engineering.

For large Doppler frequency offset, the traditional frequency offset resistant algorithm usually adopts FFT operation and spectral peak search on a received training sequence, the complexity of the algorithm is relatively high, and the frequency offset estimation precision is limited by the number of Fourier transform points.

Chinese patent publication No. CN201910908165.1 discloses a large frequency offset correction method for GMSK demodulator in AIS system. The scheme provided by the patent is only directed at the condition of frequency offset within +/-10%, but does not provide a solution for large frequency offset of +/-42%, and the frequency offset estimation scheme provided by the patent needs 4-power nonlinear operation, N-point FFT operation and spectral peak search on a baseband signal, so that the operation complexity is high.

Disclosure of Invention

The invention aims to provide a signal detection algorithm combining an improved correlation detection method and characteristic matching aiming at an AIS receiver system, solve the problem of secondary peak interference caused by an AIS training sequence non-pseudo-random sequence, fully utilize the gain of a secondary peak to achieve the detection effect closer to a pseudo-random sequence, and provide a low-complexity frequency offset estimation algorithm by utilizing the special structure of a training sequence, wherein the frequency offset estimation algorithm has high precision and wide range and can meet the requirement of a frequency offset resistance index of-4 kHz to +4 kHz.

In order to achieve the above object, the present invention provides a signal detection and frequency offset estimation algorithm based on a satellite-borne AIS system, which includes the following steps:

step one, converting an analog input signal into a digital signal after AD sampling, passing through a data delayer, directly performing quadrature frequency mixing after AD sampling if no signal is detected, and otherwise performing quadrature frequency mixing on the signal after data delay;

secondly, the mixed signal passes through a low-pass filter to filter out-of-band noise and image frequency spectrum, and the bandwidth of the filter contains the signal at the time of maximum frequency offset;

step three, carrying out data differential operation, dividing the differential signals into three paths, and respectively sending the three paths of signals to a relevant detector, a frequency offset estimation module and a matched filtering module;

step four, when the output of the correlation detector meets the peak value and is larger than the adaptive threshold, the characteristic matching is carried out, and only when the characteristic error is smaller than the set threshold, the signal is detected;

step five, if the signal is detected, finding a frame header, and performing three operations, namely performing frequency offset estimation by using the cached differential data and adjusting DDS frequency control words to perform frequency compensation according to an estimation result; secondly, performing initial phase compensation; thirdly, re-entering the training sequence and the frame header which have finished the frequency offset compensation, so that the frequency is compensated and the bit synchronization is locked in the time before the next frame header arrives;

and step six, performing matched filtering, bit synchronization, data judgment, frame head and frame tail detection and CRC (cyclic redundancy check) operation after difference.

In the third step, before the correlation detection, the mean value of the differentiated received sequence needs to be subtracted, that is, the direct current component is subtracted, and then the correlation is performed with the local sequence.

In the fourth step, a structure of constant false alarm rate detection in the radar field is adopted to obtain the adaptive threshold.

In the fourth step, it is assumed that L secondary peaks in front of the main peak are considered, and the ratios of the secondary peaks to the main peak are λ_i0Wherein i is 1, 2,. L; when the peak value is detected and is larger than the adaptive threshold, calculating the ratio lambda of the secondary peak value and the primary peak value of the fixed position before the peak value_iThen, the sum of the squared errors is calculated according to the following formula:

if e is less than delta and delta is a set threshold, the characteristic is considered to be met, and a signal is detected;

if e > δ, the signature is deemed not to be satisfied, indicating that no signal has been detected.

In the above signal detection and frequency offset estimation algorithm based on the satellite-borne AIS system, in the fifth step, the initial phase compensation includes adjusting a phase of a bit synchronization loop.

In the fifth step, the training sequence and the frame header which have completed the frequency offset compensation are re-entered, that is, the data delay switch is switched, so that the data input to the quadrature mixer is data delayed by a plurality of symbols, and the frequency is compensated and the bit synchronization is locked in the time before the next frame header arrives.

Compared with the prior art, the invention has the technical beneficial effects that:

1) and difference is firstly carried out before correlation detection, and compared with the traditional method of directly carrying out correlation detection on the received signal, the correlation peak value is not reduced or even has no peak value along with the increase of frequency offset.

2) Before the correlation detection, the mean value of the receiving sequence after the difference is required to be subtracted, namely the direct current component is subtracted, and then the receiving sequence is correlated with the local sequence, so that the correlation peak value and the peak value characteristic are almost irrelevant to the frequency offset in the frequency offset index requirement range, and a unified detector structure can be used under different frequency offset conditions.

3) The self-adaptive threshold correlation detection method is combined with the peak value feature matching method, and the secondary peak is screened out due to feature mismatch. The two are combined to overcome the problem of secondary peak interference caused by non-pseudo random codes of the training sequence, and the optimal peak position is accurately detected.

4) The characteristic that the training sequence is a 01 alternating sequence is fully utilized to carry out frequency offset estimation and compensation, the cached differential sequence is only required to be averaged at the signal detection point as the adjustment quantity of the frequency control word, and the frequency offset value can be accurately calculated under the condition of no noise, so that the method belongs to unbiased estimation, and compared with an FFT (fast Fourier transform) method, an adjacent point differential method and a multi-path correlator detection method which are commonly used in the traditional frequency offset estimation, the calculation complexity is greatly reduced.

5) The problem of large Doppler frequency offset under the satellite-borne condition can be effectively solved, the false-alarm-missing probability is far lower than the system packet loss rate within the range of-4 kHz to +4kHz required by indexes, and the system packet loss rate has no obvious relation with the frequency offset.

Drawings

The invention provides a signal detection and frequency offset estimation algorithm based on a satellite-borne AIS system, which is given by the following embodiments and attached drawings.

FIG. 1 is a block diagram of an improved AIS demodulation system implementation;

FIG. 2 is a block diagram of an improved correlation detector implementation;

FIG. 3 is a diagram of the relationship of the signal detection correlator output to the adaptive threshold;

FIG. 4 is a baseband signal;

fig. 5 is a graph of packet loss probability and false alarm probability under different frequency offset conditions.

Detailed Description

The signal detection and frequency offset estimation algorithm based on the satellite-borne AIS system according to the present invention will be described in further detail with reference to the accompanying drawings.

1) AIS demodulation system block diagram and implementation steps

The structural block diagram of the AIS demodulation system designed by the invention is shown in figure 1. The method comprises the following steps:

1) analog input signals are converted into digital signals after being AD sampled, the digital signals pass through a data delayer with a certain depth, if no signals are detected, orthogonal frequency mixing is directly carried out after the AD sampling, otherwise, the signals after the data delay are subjected to orthogonal frequency mixing, and the frequency mixing is used for moving intermediate frequency signals to zero intermediate frequency.

2) The mixed signal passes through a low-pass filter to filter out-of-band noise and image spectrum, and the bandwidth of the filter should include the signal at the time of maximum frequency offset.

3) And then carrying out data differential operation, dividing the differential signal into three paths, and respectively sending the three paths of signals to a correlation detector, a frequency offset estimation module and a matched filtering module.

4) When the output of the correlator meets the peak value and is larger than the adaptive threshold, the characteristic matching is carried out, and only when the characteristic error is smaller than the set threshold, the signal is detected.

5) If the signal is detected, finding a frame header, performing three operations, namely performing frequency offset estimation by using the cached differential data, and adjusting DDS frequency control words according to an estimation result to perform frequency compensation; adjusting the phase of the bit synchronizing ring, namely performing initial phase compensation; and thirdly, switching a data delay switch to enable the data input into the orthogonal mixer to be data delayed by a plurality of symbols, namely re-entering the training sequence and the frame header which are subjected to frequency offset compensation, so that the frequency is compensated and enough time is provided for bit synchronization locking in the period before the next frame header arrives.

6) And performing operations such as matched filtering, bit synchronization, data judgment, frame head and frame tail detection, CRC (cyclic redundancy check) and the like after the difference.

The key modules of signal detection and frequency offset estimation proposed by the present invention are specifically analyzed and described below.

2) Digital down converter

The AIS system adopts GMSK modulation, and the mathematical form of the receiver sampling signal is as follows:

wherein f is_cCarrier center frequency, f (tau) is the frequency offset introduced by the data information modulation,is used as an initial phase of the reaction,_n0for noise, the noise is not considered for the moment in order to facilitate subsequent theoretical analysis.

Due to the doppler shift introduced by the high-speed relative motion of the satellite and the transmitter and the different frequencies of the receiver and the transmitter, there exists a frequency residual Δ f in the digitally down-converted signal, which can be expressed as follows:

in discrete form as

3) Correlation detector

The starting point of the training sequence is searched by adopting a correlation detection method, and in order to improve the performance of the correlation detection, the 8-symbol frame header sequence is also utilized, and 32 symbols are added with the training sequence to carry out the correlation detection.

Before the correlation detection, the phase difference (1bit difference) is carried out on s (n) to obtain:

wherein angle (DEG) represents argument of complex number, and is realized by using cordic IP core in FPGA code, N_bRepresenting the number of sample points in one symbol period.

The phase difference signal shown in formula (4) includes doppler frequency shift Δ f and data modulation frequency shift f (n), where the doppler frequency shift Δ f is a pairThe contribution of (1) is only a direct current component, and f (n) is data modulation frequency offset, the probability of 01 is equal for the training sequence, so the contribution of the direct current component is close to 0, the data is represented as an alternating current component introduced by data alternation, the differential sequence is subtracted by the mean value thereof, namely, the direct current component is subtracted, namely, the influence of the frequency offset on the correlation value is eliminated, then the differential sequence is passed through a matched filter with the coefficient of ideal no frequency, and when the received differential sequence is matched with the local training sequence, the matched filter outputs a peak value. The self-adaptive threshold is obtained by adopting a constant false alarm rate detection structure in the radar field, and the correlation detection is considered to be effective when the peak value output by the correlator is higher than the self-adaptive threshold.

The correlation detector structure is shown in fig. 2. After correlation, the correlation result is sent to a buffer with a length of 2R +1, and assuming that the current value required to be subjected to threshold decision is stored in a ZR unit (measured unit), the adaptive threshold value at this time is calculated by the values of R reference units before the measured unit and R reference units after the measured unit, and the base power μ is obtained by taking the mean value. And multiplying the obtained base power by a threshold factor p to obtain the self-adaptive threshold value.

4) Peak feature matching

The training sequence for correlation detection is generally required to satisfy the pseudo random property. The AIS training sequence is a 01 alternating sequence, does not meet the pseudo-random characteristic, outputs a maximum peak value when being completely matched, but necessarily has slightly smaller secondary peaks near the maximum peak value, and the secondary peaks may exceed the adaptive threshold under the influence of frequency offset and noise, so that interference is caused to the judgment of a main peak. In order to solve the problem, the invention provides a feature matching method, secondary peaks are removed because the secondary peaks do not meet the features, and only main peaks are left, so that the effect of utilizing the gain of the secondary peaks is achieved.

The 01 alternating sequence becomes 0011 alternating sequence after NRZI coding, the period is 2 symbols, so that secondary peaks appear every 2n symbols on both sides of the maximum peak value, while zero points appear every 2n +1 symbols, and if no noise exists, the ratio of the secondary peaks to the main peak is constant and can be calculated in advance. Under the condition of noise, certain error exists, as long as the sum of squares of the error of the ratio of each secondary peak to the main peak and the ideal ratio is less than a certain threshold, the characteristic is considered to be met, otherwise, the characteristic is not met.

Assuming that L secondary peaks in front of the main peak are considered, ideally, the ratio of the secondary peak to the main peak is λ_i0(i ═ 1, 2.. L), when a peak is detected and greater than the adaptive threshold, then a ratio λ of the secondary peak to the primary peak is calculated at a fixed location prior to the peak_iThen, the sum of the squares of the errors is calculated according to equation (7).

If e is less than delta (delta is a set threshold), the characteristic is considered to be satisfied, and a signal is detected;

if e > δ, the signature is deemed not to be satisfied, indicating that no signal has been detected.

5) Frequency offset estimation

The term preceding equation (4) is a term related to Doppler frequency offset and is expressed asOverall offset is larger as frequency offset is larger, the latter item is relevant to the transmitted data, and if the probabilities of the transmitted data 01 are equal, the expectation of the statistical mean value is 0.

Summing the two sides of the above equation:

since the training sequence of AIS consists of 01 alternating sequences, the sum of the training sequence intervals on the right of the above formula is 0, i.e.

After substituting formula (6) to eliminate right item

Frequency control word increment

As can be seen from equation (9), buffering is only required after the signal is detectedAnd obtaining the frequency control word increment by taking the average value, and then compensating the frequency control word input by the DDS to finish the frequency offset correction.

For example: the AIS symbol rate fd 9600Hz and the simulated sampling rate 64 times the symbol rate, i.e. 614.4kHz, with a signal-to-noise ratio Eb/N0 of 12dB and a frequency offset of 4000Hz, the correlator output and the adaptive threshold are shown in fig. 3. When no signal is coming, i.e. pure noise, the correlator output value is small, much smaller than with modulated signal. And when the received training sequence is completely aligned with the local matching sequence, outputting a maximum peak value, and meanwhile, generating periodic secondary peaks around the maximum peak value, wherein the secondary peaks possibly exceed the self-adaptive threshold but can be removed by the characteristic detection module, and finally, retaining the main peak to finish frame header synchronization. After the signal is detected, the training sequence and the frame header are reentered by switching the delay switch, and the peak value is detected for the second time after N symbols, so that frequency offset estimation and compensation are not needed, bit synchronization is locked, and only the demodulation data is needed to judge the frame start mark.

As shown in fig. 4, before the signal is detected, due to the presence of frequency offset, the differentiated baseband signal has integral direct current offset, after the signal is detected and the frequency offset is estimated and compensated, the mean value of the baseband signal is corrected to be close to 0, and at this time, the data is demodulated by taking 0 as a threshold. The circles in the figure represent bit synchronization sampling points and the dotted lines represent decision thresholds.

One of the indexes for measuring the system performance is the system packet loss rate, which is mainly determined by two factors, one is caused by the false alarm of the signal detection module, and the other is caused by the error code generated by data demodulation, i.e. the signal detection is required to be correct, and about 200 data including the frame head, the frame tail, the data and the CRC check are all correctly demodulated to indicate that no packet is lost. As shown in fig. 5, the upper three curves represent the total packet loss rate of the system under the conditions of frequency offset of 0Hz, 2kHz and 4kHz, respectively. Compared with the three curves, the system packet loss rate has no obvious relation with the frequency offset. The lower three curves respectively represent the false alarm probability of the signal detection module under the conditions of frequency offset of 0Hz, 2kHz and 4 kHz. The larger the frequency offset is, the higher the false alarm probability is, but the false alarm probability is far smaller than the system packet loss probability, and the application requirements are met.