WO2010028537A1

WO2010028537A1 - Method for detecting signals of random access channel

Info

Publication number: WO2010028537A1
Application number: PCT/CN2009/000902
Authority: WO
Inventors: 薛妍; 谭欢喜; 翟羽佳
Original assignee: 中兴通讯股份有限公司
Priority date: 2008-09-10
Filing date: 2009-08-07
Publication date: 2010-03-18
Also published as: CN101355383B; CN101355383A

Abstract

The present invention discloses a method for detecting the signals of the random access channel (RACH), which includes: A. after a base station obtains N time domain correlation values R(m) of the uth root sequence, all m values corresponding to the N time domain correlation values are set as detection points, wherein m=1,2, ,N; B. among all the present detection points, the detection point m_max corresponding to the maximum time domain correlation value R(m_max) is found out; when the maximum time domain correlation value R(m_max) is greater than the peak detection threshold, it is determined that the preamble of the RACH is detected at a cyclic shift searching window corresponding to the m_max, and the m_max and the point m1 and/or point m2 of the present detection points are set as non-detection points, wherein m1=(N+m_max-d u )mod N, m2=(m_max+d u )mod N, and m1 and m2 are integer; and C. if the detection points remain, and the maximum time domain correlation value of the remained detection points is greater than the peak detection threshold, the step is jumped to step B, otherwise the detecting of the present root sequence is ended, wherein N=N zc , N zc is the length of the root sequence, d u is the span between the starting point of the cyclic shift searching window and the duplicated searching window.

Description

Signal detection method for random access channel

Technical field

The present invention relates to the field of communications, and in particular, to a signal detection method for a random access channel. Background technique

In the LTE (Long Term Evolution) system, the UE (User Equipment, User Terminal) first performs downlink synchronization through the SCH (Synchronization Channel) to find the receiving start point and the cell number of the radio frame and the subframe ( Cell ID); Then, the system information is obtained through the BCH (Broadcast Channel), and the system information includes the configuration information of the RACH (Random Access Channel). Finally, the uplink synchronization is performed through the RACH to complete the work of the access system. .

In the process of uplink synchronization, the UE first finds the transmission location of the RACH channel based on the radio frame and the reception start point of the subframe determined in the following row synchronization, and determines the available sequence of the cell for RACH transmission from the obtained system information, and then A random one of the available sequences is selected for transmission as a preamble. The base station detects the preamble to determine the amount of uplink timing adjustment and sends it to the UE. The UE adjusts the transmission timing of the uplink signal according to the timing adjustment amount, and implements uplink time synchronization.

The uplink random access preamble of LTE uses a ZC (Zadoff-Chu) sequence, and the w-th ZC sequence is defined as:

Where j is an imaginary unit; u is the root sequence index number; N _zc is the length of the ZC sequence and N _zc is a prime number, which is specified as 839 in LTE.

In LTE, each cell allocates 64 sequences for preambles, which may be different cyclic shift sequences from the same root sequence, or different cyclic shift sequences from different root sequences.

The ZC sequence is Constant Amplitude Zero Auto-correlation Code (Constant Amplitude Zero Auto-correlation Code, For the sequence of CAZAC, the correlation of ZC sequences has the following characteristics: The correlation between different cyclic shift sequences of the same sequence is 0; the correlation between different root sequences and their cyclic shift sequences is

That is, the correlation between different root sequences is very small, approximately 0. Therefore, the base station can perform time domain correlation detection on the random access signal by using the correlation property of the ZC sequence to obtain an uplink timing adjustment amount.

The time domain correlation detection method is intuitively defined as point multiplication of the received signal and the complex conjugate of each cyclic shift of the local sequence and summing to obtain the time domain correlation value of each cyclic shift sample, which can be mathematically Equivalent to the received frequency domain signal and the local frequency domain sequence complex conjugate point multiplication and then converted to the time domain by inverse Fourier transform. The mathematical form of time domain correlation detection is expressed as follows:

Suppose the time domain of the received signal is ( ), the frequency domain is ί ^); the local sequence time domain is x( ), the frequency domain is; Π^) , and the local sequence complex conjugate is in the time domain ( ) , the frequency domain form; the correlation function R(m) of the two, expressed as:

R{m) = ₍₁₎

Where m is the cyclic shift sample and Ν is the number of samples of the ZC sequence. Therefore, for a RACH user using a different cyclic shift of the same root sequence as a preamble, the received signal is converted to the frequency domain and then multiplied by the complex conjugate point of the frequency sequence of the root sequence. The result is converted into the time domain by inverse Fourier transform, and the time domain correlation value corresponding to each cyclic shift sample can be obtained.

By performing peak detection on the time domain correlation value of the cyclic shift search window of each cyclic shift sequence of the local root sequence, it is possible to know which preamble the UE uses, and obtain the UE according to the detected preamble used by the UE. Timing advance.

FIG. 1 is a schematic diagram of a RACH time domain correlation detection method. The two receiving antennas are taken as an example to describe an implementation method for time domain correlation detection of a received signal, which is briefly described as follows:

The base station first performs frequency correction and downlink sampling on the received signals of each receiving antenna, and then performs an FFT (Fast Fourier Transform) of M points (such as M=1024) to extract a RACH frequency domain signal of 839 points; Multiply the 839-point RACH frequency domain signal with the complex conjugate point of each local frequency domain root sequence, and add zero to the defect point (Ν is greater than 839, for example, Ν can be equal to 2048), and perform IFFT (Inverse Fast Fourier Transform, Fast Fourier transform The N time-domain correlation values of the local root sequence are modulo squared for the N time-domain correlation values, and the obtained N-point time-domain correlation value reflects the magnitude of the signal and the noise power; and finally the time of the two receiving antennas The domain correlation values are combined to obtain N time-domain correlation values after the antennas are combined, and then peak detection is performed to obtain the position of the preamble and the corresponding timing advance.

The performance of RACH time domain correlation detection (ie, detection of preamble, also known as RACH preamble detection) can be characterized by the missed detection rate, false alarm rate, and detection threshold of the preamble. In the RACH detection, the detection threshold corresponding to the given false alarm rate target is first determined when the signal is not transmitted, and then the detection rate of the random access signal detection when transmitting the signal is tested according to the threshold.

The false alarm rate is defined as the probability that the preamble is detected when there is no preamble transmission; the miss detection rate is defined as detecting an erroneous preamble, or not detecting a transmitted preamble, or detecting the correct preamble However, the probability of occurrence of an error timing adjustment amount or the like is estimated.

The target false alarm rate is generally required to be 10 or slightly less than 10. False alarm rate is reached when no transmission signal 10_ ³ or slightly smaller than the detection limit of the door ^10-3 of the peak detection threshold.

The threshold for peak detection is divided into absolute threshold and relative threshold. In general, the absolute threshold is related to the magnitude of the noise. When the ratio of the correlation value of the random access signal to the noise power is used as the threshold, the threshold is independent of the noise power, so the ratio of the signal correlation value to the noise power is used. Defined as the relative threshold for peak detection. When detecting the RACH signal, the system generally presets a relative threshold, and then estimates the magnitude of the noise in the detection. The noise level and the relative threshold multiply can obtain an absolute threshold, so the above is given when the signal is not sent. Determining the corresponding detection threshold under the target of the false alarm rate refers to the relative threshold. When the ratio of the correlation value of the received signal to the noise power is greater than the relative threshold, the RACH signal is considered to be present. Under normal circumstances, the higher the relative threshold is, the harder it is for users to access. The lower the setting, the more false alarms. In general, this threshold is taken as: The ratio of the signal to the noise at a false alarm rate of 10 (and therefore also the false alarm threshold). In the false alarm threshold, when the UE transmits RACH signal, if no frequency offset (the offset) and other interference, the false alarm rate is generally not higher than ^10-3. However, when there is a frequency offset, the energy of the RACH signal will be dispersed, because the preamble uses a ZC sequence, which is defined so that the energy of the RACH signal is dispersed to another cyclic shift sequence when there is a frequency offset, and thus appears. Detection error. The frequency offset characteristics of the ZC sequence are described in detail below.

After the downlink frequency offset correction, the frequency offset of the uplink RACH is not particularly large. Due to the movement of the UE The speed is generally less than 375 km/h, so it is generally considered that RACH only has a frequency offset of less than or equal to 1 time. For ease of analysis, we assume that a 1x offset (which can be a positive 1x offset or a negative 1x offset) is generated. Let the RACH sequence of the received frequency domain be ZW, and its time domain form is recorded as follows. According to the nature of IDFT (Inverse Discrete Fourier Transform), for the IDFT of N points, we have:

(n) = IDFT(z(k ± 1)) = z(«)-exp

π

According to the expression of the ZC sequence: z(«) = exp -J -un(n + 1) , the length of the sequence is L

N.

After the ZC sequence is cyclically shifted by an integer, we have to

.2π

Then z(n + d _u ) = ζ{ή) exp - / ud„ · η β (4)

J Ν "

It can be seen from the above analysis that only when ud _u =mN _zc ±\, there is z{n + d _u ) = z {n)- , that is, when the 1x frequency offset is generated, the correlation peak of the RACH sequence The offset is generated, that is, the signal in the entire cyclic shift search window is offset. Where m is an integer, and the value of m is guaranteed to be ί ·ί/„

When established, d can be expressed as: d = double the frequency offset, w · d: =m ⁺ -N _zc +\ , _m ⁺ and d _u ⁺ are integers; ud _u ~ =m~ -N _Zc -\ , -- and -- are integers. Since w.(N _zc — —) = 0— m— ).N _zc +l, SLu-nf is an integer. ^ 0<d _u ⁺ _< ^N ₂ , let N _ze - dT„ be a positive integer, N _zc /2<d- _u =N _zc -d _u ⁺ _< N _zc when < d: < N _zc , let N _zc - When d~ _u is a positive integer, then 0 < d- _u = N _zc -d: < N _zc /2. Note that N _zc is a prime number, so N _ze /2 is not an integer, and the sum is a positive integer.

In addition, observing the following formula (Equation (5)), we can also find the law of the value of positive and negative doubling.

(m- N _zc ± 1) · ") = Qxp(±j --n)

It can be seen from the above formula that if the frequency offset is a fractional multiple of the subcarrier spacing, since _Μ · will not be a decimal, the received time domain sequence will not produce a cyclic shift, but will be in the original sequence (no frequency offset). The sequence between the sequence and the sequence of the original sequence after cyclic shift (the sequence obtained after the original sequence is shifted by 1 frequency), and the received signal and the original sequence without frequency offset and 1 The sequence generated by the octave offset (that is, the sequence after the cyclic shift of the original sequence) will produce a correlation peak, and two copies of the cyclic shift search window will appear at the same time (called a replica search due to frequency offset). Window, or simply a copy window).

In 3GPP's physical layer standard TS36.211 for LTE, the values of the pair are clearly defined:

"

, where M- ' modNzc denotes Μ · = . N _ze ± l , m is the smallest integer that makes a positive integer, always less than N _Z£ /2. An example of a cyclic shift of 0 is given in Fig. 2, from which the relationship of d:, and the relationship between the cyclic shift window and the replica window can be understood, wherein the dotted line portion is an equivalent negative frequency offset window, and The negative frequency offset window used by the time domain correlation method is a cyclic cyclic shift relationship. Figure 3 shows an example of a cyclic shift window and a copy window corresponding to a cyclic shift that is not zero. As can be seen from Figures 2 and 3, the distance between the two replica windows and their cyclic shift windows can be derived using only the values defined in TS 36.211 (i.e., values less than N _zc /2). In fact, it is the distance between the start of the cyclic shift search window and the start of the copy window.

In the LTE physical layer protocol TS36.211, in LTE, the cyclic shift used for generating the preamble sequence is divided into an unrestricted set unrestricted set (also called a regular set normal set) and a restricted set restricted set (also called a high speed set high speed set). Set ) Two. The limit set (high speed set) takes into account the frequency offset effect caused by the high speed, and in order to ensure that the search windows of the respective preamble sequences do not overlap each other, the cyclic shift amount _Cv used for generating the preamble sequence is limited. However, for the non-limiting set of LTE (regular set), for scenes with small frequency offset (such as low- and medium-speed cells), the cyclic shift amount generated by the preamble sequence is not limited, and the value of the leading cyclic shift is generated. C _v = v. N _ra , where v=0, 1 , 2,

- 1 ; W means take the largest integer less than or equal to x, that is, round down X, and N _cs is the length of the search window (that is, the step size of the cyclic shift sequence).

According to the frequency offset characteristic of the ZC sequence, since the RACH signal received by the base station is frequency offset, in addition to the cyclic shift corresponding search window (referred to as a cyclic shift search window) in the RACH preamble detection, There will also be peaks in the copy window. For high-speed cells, the Doppler frequency is relatively large, so the copy window

The correlation value of the RACH signal will be relatively large. It is considered that the cyclic shift window and the two duplicate windows may have signals at the time of detection. It is necessary to comprehensively detect the cyclic shift window and the two replica windows, because the high-speed cell uses the limit set, each The cyclic shift window corresponding to each preamble and its copy window do not coincide with the cyclic shift window of other preambles and its copy window, so that there is no phenomenon that two different preambles are in one search window. For medium and low-speed cells, the Doppler frequency offset is relatively small, and the signal correlation value in the cyclic shift window is often much higher than the signal correlation value in the replica window. Therefore, only the signal in the cyclic shift window is considered during the detection, that is, only Preamble detection is performed in the cyclic shift window, and the replica window is not detected.

However, there are always certain frequency offsets in the low- and medium-speed cells, such as Doppler frequency offset. In this way, there will always be a correlation peak of the signal in the replica window, and the medium-low speed cell uses an unrestricted set, that is, there is no cyclic shift. Bit limit, although the cyclic shift window corresponding to each preamble does not coincide with the cyclic shift window of other preambles, the cyclic shift window corresponding to each preamble may coincide with the copy window of other preambles. In the low- and medium-speed cells, the signal correlation value in the cyclic shift window is much larger than the signal correlation value due to the frequency offset in the replica window, but when the signal-to-noise ratio is high, the correlation value brought by the frequency offset in the replica window It will be big. For example, only the preamble C1 is transmitted, and the preamble C2 is not transmitted. When the copy window of the current guide code C1 overlaps with the cyclic shift search window of the preamble C2, the frequency offset copy signal of C1 may fall to the cyclic shift search window of C2. When detecting the cyclic shift window of C2, it is easy to judge the frequency offset copy signal of the preamble C1 as the C2 signal, and C2 is actually not transmitted, which causes a false alarm.

This false alarm phenomenon is obvious when a RACH signal has a frequency offset and the signal to noise ratio is high. When multiple high SNR users access simultaneously, the false alarm rate will be very high. A large number of false alarms increase the signaling overhead during the random access process and increase the processing load of the system. Therefore, the false alarm needs to be suppressed.

Summary of the invention

The technical problem to be solved by the present invention is to overcome the deficiencies of the prior art and provide a signal detection method for a random access channel, which suppresses false alarms caused by frequency offsets of cells using non-limiting sets (middle and low speed cells).

In order to solve the above problems, the present invention provides a signal detection method for a random access channel, including:

A: After the base station obtains N time-domain correlation values R(m) of the u-th root sequence, N time-domain correlations are performed. All m values corresponding to the values are used as detection points, where m=l, 2, . . . , N;

B: In all current detection points, find the detection point m_max corresponding to the largest time domain correlation value R(m—max), when the maximum time domain correlation value R(m−max) is greater than the peak detection gate For a time limit, it is determined that the preamble of the random access channel is detected in the cyclic shift search window corresponding to m_max, and m_max, and the point ml and/or the point m2 in the current detection point are regarded as non-detection points; , ml = (N+m - max - d _u ) mod N, m2 = (m - max + d _u ) mod N, ml and m2 are both positive integers;

C: If there are remaining detection points, and the largest time domain correlation value in the remaining detection points is greater than the peak detection threshold, then jump to step B; otherwise, the detection of the current root sequence is ended;

Where N > N _ze , N _zc is the length of the root sequence; d _u is the distance between the start of the cyclic shift search window and the start of the replica window.

In addition, in step B, after finding the m-max, the following operations can also be performed:

If m3 > m4, a point greater than or equal to m3 and less than or equal to N, and/or a point greater than or equal to 1 and less than or equal to m4 is regarded as a non-detection point; otherwise, a point greater than or equal to m3 and less than or equal to m4 is regarded as non-detection Point; and/or

If m5>m6, a point greater than or equal to m5 and less than or equal to N, and/or a point greater than or equal to 1 and less than or equal to m6 is regarded as a non-detection point; otherwise, a point greater than or equal to m5 and less than or equal to m6 is regarded as non-detection Point

Where m3 = (N+m max - d _u - X) mod N, m4 = (N+m - max - d _u + X) mod N, m3 and m4 are positive integers; and m5 = (N+m - Max + d _u - X) mod N, m6 = (m - max + d _u + X) mod N, m5 and m6 are positive integers; X = int(N/(N _zc x 2)) , int means up or Round down.

In addition, in step B, after the m_max is found, other detection points exceeding the peak detection threshold that belong to the same cyclic shift search window as the m-max may be found in the current detection point: m[l] m [K]; identifies m[k] - d _u and/or m[k] + d _u in the current detection point as non-detection points;

1 k K.

In addition, in step B, after the detection point is found: m[l] m[K], the following operations can also be performed: If m7[k] > m8[k] , a point greater than or equal to m7[k] and less than or equal to N, and/or a point greater than or equal to 1 and less than or equal to m8[k] is used as a non-detection point; otherwise, a point greater than or equal to m7[k] and less than or equal to m8[k] as a non-detection point; and/or

If m9[k] > ml0[k] , a point greater than or equal to m9[k] and less than or equal to N, and/or a point greater than or equal to 1 and less than or equal to ml0[k] is used as the non-detection point; otherwise, a point greater than or equal to m9[k] and less than or equal to ml0[k] as a non-detection point;

Where m7[k] = (N+m[k] - d _u - X) mod N, m8[k] = (N+m[k] - d _u + X) mod N, m7[k] and m8 [k] is a positive integer; and m9[k] = (N+m[k] + d _u — X) mod N, ml0[k] = (m[k] + d _u + X) mod N, m9 [ k] and mlO [k] are positive integers; X=int(N/(N _zc x 2)) , int means rounded up or down.

In addition, in step B, after the m_max is found, all the detection points in the current detection point that belong to the same cyclic shift search window as the m-max may also be identified as non-detection points.

Further, the d _{u is} greater than or equal to 2 and less than N /2 , and when N=N _ZC , the d _u satisfies the following equation: 0 ' ) mod _zc = ±1. Further, when N>N _ZC , d _u =int( d χ N/N _zc ), and the following equation is satisfied:

O - JmodA^ = ±1 , greater than or equal to 2 and less than N /2 ; where int() means round up or round down.

In addition, the peak detection threshold can be determined by the following steps:

Performing noise estimation on the received signal to obtain a noise mean value; using the product of the above noise mean value and a preset relative threshold value as a peak detection threshold;

The relative threshold is a ratio of a false alarm signal to a noise that reaches a target false alarm rate when no signal is transmitted.

In addition, the noise mean can be determined by the following steps:

Taking the product of the maximum time domain correlation value of the received signal and the noise estimation ratio as the threshold A; taking the mean value of the time domain correlation values smaller than the threshold A among the N time domain correlation values as the noise mean value; Less than 1 and greater than 0.

Further, the root sequence is a ZC sequence. BRIEF abstract

1 is a schematic diagram of a method for detecting RACH time domain correlation;

2 is a schematic diagram showing the relationship between the positive and negative frequency offset window and the cyclic shift window when the cyclic shift is 0 in the RACH time domain correlation detection;

3 is a schematic diagram of a cyclic shift search window and a frequency offset copy window corresponding to a cyclic shift that is not 0 in RACH time domain correlation detection;

4 is a flowchart of a RACH time domain correlation detecting method according to an embodiment of the present invention;

Figure 5 is a schematic diagram of the false alarm caused by the frequency offset in the RACH correlation detection.

Preferred embodiment of the invention

The basic idea of the present invention is that, since the frequency offset causes a large number of false alarms, the position of the false alarms can be found according to the frequency offset characteristic of the ZC sequence when performing peak detection in the RACH time domain correlation detection process, and these possibilities are removed. It is the location of false alarms to achieve the purpose of false alarm suppression.

The invention will now be described in detail in conjunction with the drawings and embodiments.

4 is a flow chart of a RACH time domain correlation detection method according to an embodiment of the present invention. In order to highlight the gist of the present invention, FIG. 4 partially merges and simplifies FIG. 1, and only describes a single receiving antenna as an example. As shown in FIG. 4, the method includes the following steps:

401: extract RACH frequency domain signal values of N _zc points from the received signal;

As described above, N _zc is the length of the ZC sequence, which is specified as 839 in LTE.

402: Multiply the RACH frequency domain signal value of the N _ze points and the complex conjugate of each local root sequence to obtain N _ze point multiplication values, and add zeros to N points (ie, perform time domain over-sampling) After that, an IFFT of N points is performed, and N time-domain correlation values of each local root sequence are obtained: R(m), 1 < m <N;

Wherein, the above zero padding processing is an optional operation, and the function is that a fast inverse Fourier transform can be used to reduce the amount of calculation, and if zero padding is not performed, N = N _ZC .

403: modulo square the foregoing N time domain correlation values;

This step is an optional step. The time domain correlation value after the modulo square processing actually reflects the signal power value. The peak detection is performed on the N time-domain correlation values of each sequence. The following takes the u-th root sequence as an example, which specifically includes the following steps:

404: In the N time domain correlation values, all m values (cyclic shift samples, or detection points) corresponding to time domain correlation values greater than the peak detection threshold (denoted as threshold B) are taken as initial waiting Set of detection points;

Threshold B can be calculated as follows:

404A: Perform noise estimation on the received signal to obtain the noise threshold (denoted as threshold A); Threshold A can be equal to: Maximum time domain correlation value X noise estimation ratio; where noise estimation ratio is greater than 0 and less than 1, for example, noise estimation ratio The value can be taken as 0.6.

404B: Calculate the mean of all time-domain correlation values less than threshold A (called the noise mean); the noise mean actually reflects the amount of noise power.

Of course, the threshold A and the noise mean can also be calculated using other methods of the prior art.

The relative threshold can be the ratio of the false alarm signal to the noise at the target false alarm rate (for example, 0.1%) when no signal is sent.

It should be noted that the role of the peak detection threshold (threshold B) is to distinguish between noise and signal. In the present invention, only the function of stopping detection is performed (that is, the time domain correlation value less than the threshold is not detected), in the prior art. There are many other settings and calculation methods.

405: Find a maximum value R(m max) of the time domain correlation value of each point to be detected in the current set of points to be detected, and determine that the preamble is detected in the cyclic shift search window corresponding to m—max, and according to m—max Determining a cyclic shift amount and a timing advance of the root sequence corresponding to the preamble;

406: Remove m—max from the current set of points to be detected (to avoid repeated detection), if m−max− _du and/or m−max+d _{u are} located in the current to-be-detected set, m−max− _du And / or m - max + d _u is deleted from the current set of points to be detected;

Considering that m - max - d _u and m - max + d _u may not be in the range of interval [1 , N], the point actually deleted from the set of points to be detected is: ml = (N+m - max - d _u ) mod N, m2 = (m— max + d _u ) mod N, ml and m2 are both positive integers. In addition m- max - d _u and m- max + d _u to Xi Bu, can capture interval ^ _{[m- max- d u - X,} m- max - d u + X] and [m- max + d _u - Other values in X, m - max + d _u + X] are removed from the set of points to be detected. Where [ ] denotes a closed interval; X= int(N/(N _zc x2)), and int denotes rounding up or down.

Similarly, considering that m - max - d _u ± X and m - max + d _u ± X may not be in the range of interval [1 , N], points that satisfy the following conditions should actually be deleted from the set of points to be detected :

If m3 > m4, delete a point greater than or equal to m3 and less than or equal to N, and/or delete a point greater than or equal to 1 and less than or equal to m4; otherwise, delete a point greater than or equal to m3 and less than or equal to m4; if m5>m6, Then deleting a point greater than or equal to m5 and less than or equal to N, and/or deleting a point greater than or equal to 1 and less than or equal to m6; otherwise, deleting a point greater than or equal to m5 and less than or equal to m6; wherein, m3 = (N+m max - d _u - X) mod N, m4 = (N+m - max - d _u + X) mod N, m3 and m4 are positive integers;

M5 = (N+m - max + d _u - X) mod N, m6 = (m - max + d _u + X) mod N, m5 and m6 are positive integers.

The above d _u is the distance between the start point of the cyclic shift search window and the start of the replica window. When du=l, the loop shift window and the copy window almost coincide, and it is not necessary to delete the point of the copy window. Therefore, it is possible to limit the operation of step 406 only when d _{u is} greater than or equal to 2.

When N=N _ZC , it satisfies: 0 · ) mod N _zc = ±1 and less than N _Zc /2;

When N>N _ZC , d _u =int< „ x N/N _zc ), and 0· ί ' _Μ )mod _zc = ±1 and ^ is less than N / 2 , where int() means round up or down Rounding.

u is the root sequence number, the operator mod represents the remainder, and when 0 < xy, the value of (x) mod (y) is less than or equal to " /2 , "·] means round up, for example: 838 mod 839= - 1.

In this step, the preamble corresponding to the cyclic shift window in which the maximum correlation value in the current set of points to be detected is located is used as the detected preamble, and the position of the maximum correlation value, m_max, and the maximum correlation value are cyclically shifted. m- max positions corresponding to positions of the two copies of the window level and the window corresponding to the maximum correlation value (m- max - d _u and m- max + d _u) and the position m- max _ d _u, the position m- max + A number of positions adjacent to d _u are removed from the set of points to be detected as false alarm positions to reduce the false alarm rate.

407: If the to-be-detected point set further includes a to-be-detected point, skip to step 405, otherwise Peak detection of the uth root sequence of the beam.

Fig. 5 is a schematic diagram of a false alarm caused by frequency offset in RACH correlation detection; an application example of the present invention will be described below with reference to Fig. 5.

After obtaining the N time domain correlation values corresponding to a certain sequence, the time domain correlation values smaller than the threshold A are averaged, and the mean values of the data reflect the noise power mean of the root sequence. Means and noise false alarm rate relative to the target gate ^10-3 when the threshold limit value as the product of B. Suppose the length of the search window configured by the system is Ncs, there is a RACH user access, the user uses a cyclic shift Cvl of a certain root sequence, then the user's cyclic shift search window from Cvl to Cvl+Ncs, one The copy window is Cvl - d _u to Cvl - d _u + Ncs; the other copy window is Cvl + d _u to Cvl + d _u + Ncs. For low-speed cells, there is no cyclic shift limitation, so the replica window of the cyclic shift Cvl may be a cyclic shift search window of other RACH users.

As shown in Fig. 5, when all the detected positions are detected, it is found that the points xl (ml, R(ml)) and the points x2 (m2, R(m2)) are points exceeding the threshold B; the peak corresponding to the point xl The power R(ml) is the largest and falls within the cyclic shift search window of Cvl, so it is determined that the cyclic shift is a RACH user of Cvl. Point x2 is also the point that exceeds threshold B. In fact, point x2 is caused by the frequency offset of RACH at point xl, but in the case where the low- and medium-speed cells have no cyclic shift limit, because x2 falls within the cyclic shift window of Cv2. If false alarm suppression is not performed, point x2 will be judged as another RACH user whose cyclic shift is Cv2, and point x2 will become a false alarm. If the signal-to-noise ratio is high, a false alarm will appear in the copy window corresponding to the cyclic shift window of the RACH signal, which will bring a lot of signal overhead.

The distance between point x2 and point xl is exactly d _u , which means that point x2 may be caused by the frequency offset of point xl. Point x2 and its left and right points may be false alarm points. In the detection, point x2 and The left and right points are removed from the set of points to be detected, so that the signals at these positions are not detected, and no false alarms are generated. Since the preamble signal has been detected by the cyclic shift search window corresponding to the point x1, the non-detection position further includes all points in the cyclic shift search window corresponding to the point x1. In order to avoid repeated detection, it is also necessary to move from Cvl to Cvl. + The location of Ncs is marked as a non-detected location. Next, all the detectable positions are detected again, and it is found that the time domain correlation value of the threshold B is not exceeded, and then the peak detection process of the root sequence is determined to be ended. The final result of the correlation detection of the root sequence is that only the cyclic shift of the point xl is detected. The leading position corresponding to the bit search window, corresponding to the cyclic shift search window where the point x2 is located The lead is not reported as a result of the test.

In some rare cases, the peak point of a cyclic shift cvl frequency offset copy coincides with the same sequence and another cyclic shift cv2 peak point. In this case, the above false alarm suppression method may cause The missed detection of the signal corresponding to cv2 is cyclically shifted, but this situation rarely occurs because the position of each user in the actual system is random from the base station, and the delay is also different. And when the length of the search window is large, the available cyclic shift of the same sequence will be small, and the probability of such coincidence is even smaller.

In addition, we limit the false alarm suppression process of the present invention in the case where d _{u is} less than 2 (that is, in the peak detection, the position determined as the preamble in a certain cyclic shift search window is regarded as the non-corresponding position of the copy window thereof. The detection position) ensures that there is no missed detection when there is a peak overlap between a cyclic shift window and its own copy.

According to the basic principle of the present invention, the above embodiment has various transformations, such as:

(1) In step 405, after determining m_max, the following processing may also be performed: First, all the points in the set of points to be detected are located in the same cyclic shift search window as m-max, and then located at m-max. The points of the same cyclic shift search window and/or the corresponding points in the copy window corresponding to the cyclic shift search window are deleted from the set of points to be detected.

That is, if there is m maxl located in the same cyclic shift search window as m-max in the set of points to be detected, the points in the copy window corresponding to m maxl are m maxl - d _u and m maxl + d _u ( If m maxl, m maxl - d _u and m maxl + d _{u are} in the set of points to be detected, indicating that the corresponding time domain correlation value is greater than the threshold B), m_maxl may be deleted from the set of points to be detected. If m maxl, m maxl - d _u and / or m maxl + d _u be the set of detection points may be m maxl - d _u and / or m- maxl + d _u to be removed from the collection point detection.

Of course, the interval may also be won ^ _{[m- maxl- d u - X,} m maxl - d u + X] , and - other values _{[m- maxl + d u X,} m maxl + d u + X] to be in the The detection point is removed from the collection.

In general, after finding other detection points that exceed the peak detection threshold of m-max in the same cyclic shift search window in the current detection point: m[l] m[K], the following operations can also be performed: If m7[ k] > m8[k] , then a point greater than or equal to m7[k] and less than or equal to N, and/or a point greater than or equal to 1 and less than or equal to m8[k] as a non-detection point; otherwise, greater than or equal to m7 [k] and small a point equal to m8[k] as a non-detection point; and/or

Where m7 [k] = (N+m[k] - d _u - X) mod N , m8 [k] = (N+m[k] - d _u + X) mod N , m7[k] and m8 [k] is a positive integer;

M9[k] = (N+m[k] + d _u - X) mod N, ml0[k] = (m[k] + d _u + X) mod N, m9 [k] and mlO [k] are A positive integer.

(II) The above embodiment describes the present invention by taking a single receiving antenna as an example. The present invention is also applicable to the case of multiple receiving antennas, and only needs to add N time-domain correlation values corresponding to the receiving antennas separately. Peak detection is sufficient.

(3) In the embodiment, the set of points to be detected is used, and each point in the set of detection points is subjected to peak detection, and each non-detected point or the detected point is deleted from the set of points to be detected; Other methods can be used, for example, to identify detected or non-detected points to avoid repeated detection or detection of false alarm points.

Industrial applicability

In summary, the method of the present invention can effectively suppress false alarms caused by frequency offset in cells using non-limiting sets (middle and low-speed cells), reduce signaling overhead caused by false alarms, and avoid The missed detection of the random access signal ensures the successful access of more users.

Claims

Claim

A signal detection method for a random access channel, comprising:

A: After the base station obtains the N time domain correlation values R(m) of the uth root sequence, all the m values corresponding to the N time domain correlation values are used as detection points, where m=l, 2, .. . , N;

Where N > N _ZC , N _zc is the length of the root sequence; d _u is the distance between the start of the cyclic shift search window and the start of the replica window.

2. The method of claim 1 wherein:

In the step B, after the m-max is found,

If m3 > m4, a point greater than or equal to m3 and less than or equal to N, and/or a point greater than or equal to 1 and less than or equal to m4 is regarded as a non-detection point; otherwise, a point greater than or equal to m3 and less than or equal to m4 is regarded as a non-detection point And/or if m5>m6, a point greater than or equal to m5 and less than or equal to N, and/or a point greater than or equal to 1 and less than or equal to m6 is regarded as a non-detection point; otherwise, a point greater than or equal to m5 and less than or equal to m6 As a non-detection point;

Where m3 = (N+m max - d _u - X) mod N, m4 = (N+m - max - d _u + X) mod N, m3 and m4 are positive integers;

M5 = (N+m - max + d _u - X) mod N, m6 = (m - max + d _u + X) mod N, m5 and m6 are positive integers;

X=int (N /(N _zc x 2)) , int means round up or down.

3. The method of claim 1 wherein:

In the step B, after the m-max is found,

Finding other detection points exceeding the peak detection threshold that belong to the same cyclic shift search window as the m-max in the current detection point: m[l] m[K] ;

m[k] - d _u and/or m[k] + d _u in the current detection point as non-detection points;

Where 1 k K.

4. The method of claim 3, wherein:

In the step B, after the detection point is found: m[l] m[K],

If m7[k] > m8[k] , a point greater than or equal to m7[k] and less than or equal to N, and/or a point greater than or equal to 1 and less than or equal to m8[k] is regarded as a non-detection point; otherwise, it will be greater than a point equal to m7[k] and less than or equal to m8[k] as a non-detection point; and/or

If m9[k] > ml0[k] , a point greater than or equal to m9[k] and less than or equal to N, and/or a point greater than or equal to 1 and less than or equal to ml0[k] is regarded as a non-detection point; otherwise, it will be greater than a point equal to m9[k] and less than or equal to ml0[k] as a non-detection point;

Where m7 [k] = (N+m[k] - d _u - X) mod N , m8 [k] = (N+m[k] - d _u + X) mod N , m7[k] and m8 [k] is a positive integer; and m9[k] = (N+m[k] + d _u - X) mod N, ml0[k] = (m[k] + d _u + X) mod N, m9 [ k] and mlO [k] are positive integers;

X=int (N l(N _zc x 2)) , int means round up or down.

5. The method of claim 1 or 3 or 4, wherein:

In the step B, after the m-max is found, all the detection points in the current detection point that belong to the same cyclic shift search window as the m-max are regarded as non-detection points.

6. The method of claim 1 wherein:

The d _{u is} greater than or equal to 2 and less than N /2 , and when N=N _ZC , the d _u satisfies: ( ) mod _zc = ±l.

7. The method of claim 1 wherein:

When N>N _ZC , d _u =int( ' _a N/N _zc ), 3.d satisfies: ( · d ) moA N _zc = ±1 , d _u Greater than or equal to 2 and less than N /2 ;

Where int() means rounding up or rounding down.

8. The method of claim 1 wherein said peak detection threshold is determined using the following steps: noise estimation of the received signal to obtain a noise mean; product of said noise mean and a predetermined relative threshold As a peak detection threshold; wherein the relative threshold is a ratio of a false alarm signal to a noise that reaches a target false alarm rate when no signal is transmitted.

9. The method according to claim 8, wherein the noise mean is determined by using: a product of a maximum time domain correlation value of the received signal and a noise estimation ratio as a threshold A; the N time domain correlation values The mean value of the time domain correlation value smaller than the threshold A is taken as the noise mean value; wherein the noise estimation ratio is less than 1 and greater than 0.

10. The method of claim 1 wherein the root sequence is a ZC sequence.