Design method and system of prototype filter in cosine modulation non-uniform filter bank
Patent Information
- Application Number
- CN202410985942.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-23
- Publication Date
- 2024-11-12
Smart Images

Figure CN118940501A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to a design method and system for a filter, and in particular to a design method and system for a prototype filter in a cosine modulated non-uniform filter group. Background Art
[0002] Cosine modulated filter banks can be used to realize the overall modulation and demodulation of multi-carriers in multi-frequency-time division multiple access systems. Its complete structure includes analytical filter banks and integrated filter banks. Both can be obtained by cosine modulating the same low-pass prototype filter. The parameters of the prototype filter highly affect the performance of the system, including channel interference and symbol interference, as well as the degree of out-of-band leakage and bit error rate, and will also affect the channel estimation accuracy, prediction performance and synchronization accuracy of the system including the channel estimation module, channel prediction module and synchronization module. Therefore, it is necessary to reduce the order of the prototype filter as much as possible under the premise of ensuring system performance to adapt to various application scenarios with different data rates, different protection bandwidth widths and different implementation complexity requirements. However, uniform multi-carrier demodulation cannot realize the simultaneous demodulation processing of multiple carriers with different carrier rates and arbitrary arrangement of subcarrier center frequencies. Therefore, it is necessary to design a non-uniform filter bank based on the cosine modulated uniform filter bank structure to realize the non-uniform channelization of the system and restore the original non-uniform carrier signal without error. At present, the research on the reconstruction conditions of the non-uniform filter bank based on cosine modulation is relatively rich, but there is still less research on the design method of the prototype filter. Summary of the invention
[0003] In view of the problems existing in the prior art, the object of the present invention is to provide a design method and system for a prototype filter in a cosine modulated non-uniform filter bank with low complexity.
[0004] In order to achieve the above-mentioned object of the invention, the present invention provides the following technical solutions:
[0005] A method for designing a prototype filter in a cosine modulated non-uniform filter bank comprises the following steps:
[0006] (1) The prototype filter optimization model is established as follows:
[0007]
[0008] Where g = (g(0), g(1), …, g(N-1)) T is the coefficient vector of the prototype filter, g(i) is the i-th filter coefficient, N is the number of filter coefficients, α and β are weight factors, E1 is the reconstruction error of the non-uniform filter bank, E2 is the amplitude error of the prototype filter passband, and E3 is the amplitude error of the prototype filter stopband;
[0009] (2) converting the prototype filter optimization model into a quadratic function model through frequency sampling;
[0010] (3) Iteratively solving the quadratic function model, and taking the coefficient vector at the end of the iteration as the coefficient vector of the optimal prototype filter.
[0011] Furthermore, the reconstruction error of the non-uniform filter bank is specifically:
[0012]
[0013] Among them, G(e jω )、G(e j(ω-πM) ) are the frequency responses of the prototype filters of the cosine modulated non-uniform filter bank at frequencies ω and ω-π / M, respectively, and M is the number of channels of the cosine modulated non-uniform filter bank;
[0014] The amplitude error of the prototype filter passband is specifically:
[0015]
[0016] Among them, ω p is the passband cutoff frequency of the prototype filter, ω p =π / 2M;
[0017] The amplitude error of the prototype filter stop band is specifically:
[0018]
[0019] Among them, ω s is the stopband cutoff frequency of the prototype filter, ω s =(π+ε) / 2M, ε is the frequency spacing of adjacent channels in the filter bank.
[0020] Furthermore, the quadratic function model is specifically:
[0021]
[0022] Where p = (g(0), g(1), ..., g(N / 2-1)) T is the coefficient vector of the first half of the coefficients of the prototype filter, N is the number of filter coefficients, p (t) represents the value of p at the tth iteration, p (0) =0, α, β are weight factors respectively, K represents the number of sampling points, M is the number of channels of the cosine modulated uniform filter bank, I is the unit column vector, ω p is the passband cutoff frequency of the prototype filter, U, Q, U p , U s is an intermediate variable, and the calculation formula is:
[0023] U=H(Ω p ) r (Ω p )+H(Ω p -π / M)U r (Ω p -π / M)
[0024]
[0025] C(ω)=[cos(N-1)ω / 2,cos(N-3)ω / 2,…,cosω / 2] T
[0026]
[0027] In the formula, H(Ω p ), U r (Ω p )、H(Ω p -π / M)、U r (Ω p -π / M)、 Ω p are all intermediate variables, ω s Represents the stopband cutoff frequency of the prototype filter.
[0028] Furthermore, step (3) specifically includes:
[0029] (3.1) Solve the quadratic function model and obtain the optimal solution of the coefficient vector at the tth iteration:
[0030]
[0031] in, U q is an intermediate variable;
[0032] (3.2) According to p (t) Calculate the coefficient vector p for the next iteration (t+1) for:
[0033] p (t+1) =(1-ξ)p (t) +ξp (t-1) ,t≥1
[0034] Where ξ is the smoothing factor, 0<ξ<1;
[0035] (3.3) Determine whether the following conditions are met. If so, stop the iteration and set p (t) As the optimal solution, and p (t)According to the even symmetry characteristic, the coefficient vector of the second half coefficients is obtained, and then the first half coefficients and the second half coefficients are combined to obtain the coefficient vector of the prototype filter; otherwise, t=t+1 is set and the execution of step (3.1) is returned;
[0036] ||p (t+1) -p (t) ||2 / ||p (t) ||2<χ
[0037] Where χ represents the error limit.
[0038] Furthermore, 0<α<1, 0≤β<1, α+β≤1.
[0039] A design system for a prototype filter in a cosine modulated non-uniform filter bank, comprising:
[0040] The optimization model building module is used to build the prototype filter optimization model:
[0041]
[0042] Where g = (g(0), g(1), …, g(N-1)) T is the coefficient vector of the prototype filter, g(i) is the i-th filter coefficient, N is the number of filter coefficients, α and β are weight factors, E1 is the reconstruction error of the non-uniform filter bank, E2 is the amplitude error of the prototype filter passband, and E3 is the amplitude error of the prototype filter stopband;
[0043] A model conversion module, used for converting the prototype filter optimization model into a quadratic function model through frequency sampling;
[0044] The model solving module is used to iteratively solve the quadratic function model and use the coefficient vector at the end of the iteration as the coefficient vector of the optimal prototype filter.
[0045] Furthermore, the reconstruction error of the non-uniform filter bank is specifically:
[0046]
[0047] Among them, G(e jω )、G(e j(ω-πM) ) are the frequency responses of the prototype filters of the cosine modulated non-uniform filter bank at frequencies ω and ω-π / M, respectively, and M is the number of channels of the cosine modulated non-uniform filter bank;
[0048] The amplitude error of the prototype filter passband is specifically:
[0049]
[0050] Among them, ωp is the passband cutoff frequency of the prototype filter, ω p =π / 2M;
[0051] The amplitude error of the prototype filter stop band is specifically:
[0052]
[0053] Among them, ω s is the stopband cutoff frequency of the prototype filter, ω s =(π+ε) / 2M, ε is the frequency spacing of adjacent channels in the filter bank.
[0054] Furthermore, the quadratic function model is specifically:
[0055]
[0056] Where p = (g(0), g(1), ..., g(N / 2-1)) T is the coefficient vector of the first half of the coefficients of the prototype filter, N is the number of filter coefficients, p (t) represents the value of p at the tth iteration, p (0) =0, α, β are weight factors respectively, K represents the number of sampling points, M is the number of channels of the cosine modulated uniform filter bank, I is the unit column vector, ω p is the passband cutoff frequency of the prototype filter, U, Q, U p , U s is an intermediate variable, and the calculation formula is:
[0057] U=H(Ω p ) r (Ω p )+H(Ω p -π / M)U r (Ω p -π / M)
[0058]
[0059] C(ω)=[cos(N-1)ω / 2,cos(N-3)ω / 2,…,cosω / 2] T
[0060]
[0061] In the formula, H(Ω p ), U r (Ω p )、H(Ω p -π / M)、U r (Ω p -π / M)、 Ω p are all intermediate variables, ω s Represents the stopband cutoff frequency of the prototype filter.
[0062] Furthermore, the model solving module specifically includes:
[0063] The first solving unit is used to iteratively solve the quadratic function model, and obtain the optimal solution of the coefficient vector at the t-th iteration as:
[0064]
[0065] in, U q is an intermediate variable;
[0066] The second solving unit is used to calculate the value of p (t) Calculate the coefficient vector p for the next iteration (t+1) for:
[0067] p (t+1) =(1-ξ)p (t) +ξp (t-1) ,t≥1
[0068] Where ξ is the smoothing factor, 0<ξ<1;
[0069] The condition judgment module is used to judge whether the following conditions are met. If they are met, the iteration is stopped and p (t) As the optimal solution, and p (t) According to the even symmetry characteristic, the coefficient vector of the second half coefficients is obtained, and then the first half coefficients and the second half coefficients are combined to obtain the coefficient vector of the prototype filter; otherwise, t=t+1 is set and the first solution unit is returned to execute;
[0070] ||p (t+1) -p (t) ||2 / ||p (t) ||2<χ
[0071] Where χ represents the error limit.
[0072] Furthermore, 0<α<1, 0≤β<1, α+β≤1.
[0073] Compared with the prior art, the invention has the following beneficial effects: a prototype filter optimization model is established for the weighted average function of three errors in the non-uniform filter group, wherein the three errors are the filter group reconstruction error, the amplitude errors of the prototype filter passband and the stopband; the prototype filter optimization model is converted into a quadratic function model through frequency sampling, and then the minimum value of the quadratic function model is obtained, and the minimum value is the optimal prototype filter coefficient. Thus, the reconstruction of the filter group is realized by using a prototype filter of a lower order, the realization complexity of the filter group is reduced, and the calculation resources when the filter group is realized in hardware are reduced. BRIEF DESCRIPTION OF THE DRAWINGS
[0074] Figure 1 It is a flow chart of a method for designing a prototype filter in a cosine modulated non-uniform filter bank provided by an embodiment of the present invention;
[0075] Figure 2 It is a 18MHz non-uniform broadband signal;
[0076] Figure 3 It is the cosine modulated uniform filter bank structure of M channels. DETAILED DESCRIPTION
[0077] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention.
[0078] Embodiment 1
[0079] The embodiment of the present invention provides a method for designing a prototype filter in a cosine modulated non-uniform filter bank. Figure 1 As shown, the following steps are included:
[0080] (1) The prototype filter optimization model is established as follows:
[0081]
[0082] Where g = (g(0), g(1), …, g(N-1)) T is the coefficient vector of the prototype filter, g(i) is the i-th filter coefficient, N is the number of filter coefficients, α, β are weight factors, 0<α<1, 0≤β<1, α+β≤1, E1 is the reconstruction error of the non-uniform filter bank, E2 is the amplitude error of the prototype filter passband, and E3 is the amplitude error of the prototype filter stopband.
[0083] Specifically, the reconstruction error of the non-uniform filter bank is:
[0084]
[0085] Among them, G(e jω)、G(e j(ω-π / M) ) are the frequency responses of the prototype filters of the cosine modulated non-uniform filter bank at frequencies ω and ω-π / M, respectively, and M is the number of channels of the cosine modulated non-uniform filter bank;
[0086] The amplitude error of the prototype filter passband is specifically:
[0087]
[0088] Among them, ω p is the passband cutoff frequency of the prototype filter, ω p =π / 2M;
[0089] The amplitude error of the prototype filter stop band is specifically:
[0090]
[0091] Among them, ω s is the stopband cutoff frequency of the prototype filter, ω s =(π+ε) / 2M, ε is the frequency spacing of adjacent channels in the filter bank.
[0092] (2) The prototype filter optimization model is converted into a quadratic function model through frequency sampling.
[0093] Specifically, transform E1 into
[0094]
[0095] Where t is the number of iterations. When t = 0, the vector p (0) Initialized to 0, K represents the number of sampling points, and I is the unit column vector;
[0096] U=H(Ω p ) r (Ω p )+H(Ω p -π / M)U r (Ω p -π / M)
[0097] in,
[0098]
[0099] diag() is a diagonal matrix, p = (g(0), g(1), ..., g(N / 2-1)) T is the coefficient vector of the prototype filter;
[0100] C(ω)=[cos(N-1)ω / 2,cos(N-3)ω / 2,…,cosω / 2] T ,
[0101] Ω p is the sampling point, expressed as:
[0102]
[0103] Transform E2 into
[0104] E2=ω p -p T Q+p T U p p
[0105] in,
[0106] Transform E3 into
[0107] E3=p T U s p
[0108] Among them, U s The expression is:
[0109]
[0110] After the above transformation, the optimization model can be transformed into a quadratic function model:
[0111]
[0112] (3) Iteratively solving the quadratic function model, and taking the coefficient vector at the end of the iteration as the coefficient vector of the optimal prototype filter.
[0113] This step specifically includes:
[0114] (3.1) Solve the quadratic function model and obtain the optimal solution of the coefficient vector at the tth iteration:
[0115]
[0116] in,
[0117] (3.2) According to p (t) Calculate the coefficient vector p for the next iteration (t+1) for:
[0118] p (t+1) =(1-ξ)p (t) +ξp (t-1) ,t≥1
[0119] Where ξ is the smoothing factor, 0<ξ<1.
[0120] (3.3) Determine whether the following conditions are met. If so, stop the iteration and set p (t) As the optimal solution, and p (t) According to the even symmetry characteristic, the coefficient vector of the second half coefficients is obtained, and then the first half coefficients and the second half coefficients are combined to obtain the coefficient vector of the prototype filter; otherwise, t=t+1 is set and the execution of step (3.1) is returned;
[0121] ||p (t+1) -p (t) ||2 / ||p (t) ||2<χ
[0122] Where χ represents the error limit.
[0123] A specific application example of the present invention is as follows Figure 2 As shown, specifically, in the 18MHz non-uniform broadband signal, it consists of three sub-band signals with a bandwidth of 4MHz and three sub-band signals with a bandwidth of 2MHz. Consider a 360-channel uniform filter bank, such as Figure 3 The M-channel cosine modulated uniform filter bank structure shown is divided into two parts: the analytical filter bank and the comprehensive filter bank. Each part is composed of M-channel sub-filters, that is, M = 360, and a non-uniform filter bank is implemented on this basis. The prototype filter passband cutoff frequency is set to ω p is 0.004363323, the stop band cutoff frequency ω s is 0.0087266, the stopband attenuation is set to -40dB; the weight factors α and β are fixed weight factors, α = 0.4, β = 0.1. Set the sampling factor K to the power of 2, K = 128, and the iteration error limit value χ = 1×10 -11 , smoothing factor ξ=0.5, “guard band” width τ=1×10 -6 , allowable error τ e Set to 1×10 -11 According to the DVB-RCS2 standard, the transition band width ε=0.0611 is set, the number of input users is set to 6, and the number of sub-channels required to be merged for each sub-band signal of the non-uniform filter bank to process the broadband signal is {76,76,76,37,37,37}, that is, a total of 21 sub-channels are used as protection bands to meet the index requirements of the transition band width. The design method of the present invention is adopted, and finally the reconstruction of the filter bank is realized with a prototype filter of a lower order, and the complexity is low.
[0124] Embodiment 2
[0125] An embodiment of the present invention provides a design system for a prototype filter in a cosine modulated non-uniform filter bank, comprising:
[0126] The optimization model building module is used to build the prototype filter optimization model:
[0127]
[0128] Where g = (g(0), g(1), …, g(N-1)) T is the coefficient vector of the prototype filter, g(i) is the i-th filter coefficient, N is the number of filter coefficients, α and β are weight factors, E1 is the reconstruction error of the non-uniform filter bank, E2 is the amplitude error of the prototype filter passband, and E3 is the amplitude error of the prototype filter stopband;
[0129] A model conversion module, used for converting the prototype filter optimization model into a quadratic function model through frequency sampling;
[0130] The model solving module is used to iteratively solve the quadratic function model and use the coefficient vector at the end of the iteration as the coefficient vector of the optimal prototype filter.
[0131] The reconstruction error of the non-uniform filter bank is specifically:
[0132]
[0133] Among them, G(e jω )、G(e j(ω-π / M) ) are the frequency responses of the prototype filters of the cosine modulated non-uniform filter bank at frequencies ω and ω-π / M, respectively, and M is the number of channels of the cosine modulated non-uniform filter bank;
[0134] The amplitude error of the prototype filter passband is specifically:
[0135]
[0136] Among them, ω p is the passband cutoff frequency of the prototype filter, ω p =π / 2M;
[0137] The amplitude error of the prototype filter stop band is specifically:
[0138]
[0139] Among them, ω s is the stopband cutoff frequency of the prototype filter, ω s =(π+ε) / 2M, ε is the frequency spacing of adjacent channels in the filter bank.
[0140] Wherein, the quadratic function model is specifically:
[0141]
[0142] Where p = (g(0), g(1), ..., g(N / 2-1)) T is the coefficient vector of the prototype filter, N is the number of filter coefficients, p (t) represents the value of p at the tth iteration, p (0) =0, α, β are weight factors respectively, K represents the number of sampling points, M is the number of channels of the cosine modulated uniform filter bank, I is the unit column vector, ω p is the passband cutoff frequency of the prototype filter, U, Q, U p , U s The calculation formula is:
[0143] U=H(Ω p ) r (Ω p )+H(Ω p -π / M)U r (Ω p -π / M)
[0144]
[0145] C(ω)=[cos(N-1)ω / 2,cos(N-3)ω / 2,…,cosω / 2] T
[0146]
[0147] In the formula, ω s Represents the stopband cutoff frequency of the prototype filter.
[0148] Wherein, the model solving module specifically includes:
[0149] The first solving unit is used to solve the quadratic function model and obtain the optimal solution of the coefficient vector at the t-th iteration:
[0150]
[0151] in,
[0152] The second solving unit is used to calculate the value of p (t) Calculate the coefficient vector p for the next iteration (t+1) for:
[0153] p (t+1) =(1-ξ)p (t) +ξp (t-1) ,t≥1
[0154] Where ξ is the smoothing factor, 0<ξ<1;
[0155] The condition judgment module is used to judge whether the following conditions are met. If they are met, the iteration is stopped and p (t) As the optimal solution, and p (t) According to the even symmetry characteristic, the coefficient vector of the second half coefficients is obtained, and then the first half coefficients and the second half coefficients are combined to obtain the coefficient vector of the prototype filter; otherwise, t=t+1 is set and the first solution unit is returned to execute;
[0156] ||p (t+1) -p (t) ||2 / ||p (t) ||2<χ
[0157] Where χ represents the error limit.
[0158] The system provided in the embodiment of the present invention can be used to execute the method provided in the first embodiment of the present invention, and has the corresponding functions and beneficial effects of the execution method.
[0159] It is worth noting that in the embodiment of the above system, the various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of the functional units are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of the present invention.
[0160] The embodiments described above are only illustrative, wherein the modules described as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical modules, i.e., they may be located in one place, or they may be distributed on multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, or of course, it can be implemented only by hardware, as long as the function or effect can be achieved.
Claims
1. A method for designing a prototype filter in a cosine modulated non-uniform filter bank, characterized in that: The steps include: (1) The prototype filter optimization model is established as follows: Where g = (g(0), g(1), …, g(N-1)) T is the coefficient vector of the prototype filter, which has an even symmetric characteristic, g(i) is the i-th filter coefficient, N is the number of filter coefficients, α and β are weight factors, E1 is the reconstruction error of the non-uniform filter bank, E2 is the amplitude error of the prototype filter passband, and E3 is the amplitude error of the prototype filter stopband; (2) converting the prototype filter optimization model into a quadratic function model through frequency sampling; (3) Iteratively solving the quadratic function model, and taking the coefficient vector at the end of the iteration as the coefficient vector of the optimal prototype filter.
2. The method for designing a prototype filter in a cosine modulated non-uniform filter bank according to claim 1, wherein: The reconstruction error of the non-uniform filter bank is specifically: Among them, G(e jω )、G(e j(ω-π / M) ) are the frequency responses of the prototype filters of the cosine modulated non-uniform filter bank at frequencies ω and ω-π / M, respectively, and M is the number of channels of the cosine modulated non-uniform filter bank; The amplitude error of the prototype filter passband is specifically: Among them, ω p is the passband cutoff frequency of the prototype filter, ω p =π / 2M; The amplitude error of the prototype filter stop band is specifically: Among them, ω s is the stopband cutoff frequency of the prototype filter, ω s =(π+ε) / 2M, ε is the frequency spacing of adjacent channels in the filter bank.
3. The design method of the prototype filter in the cosine modulated non-uniform filter bank according to claim 1, characterized in that: The quadratic function model is specifically: Where p = (g(0), g(1), ..., g(N / 2-1)) T is the coefficient vector of the first half of the coefficients of the prototype filter, N is the number of filter coefficients, p (t) represents the value of p at the tth iteration, p (0) =0, α, β are weight factors respectively, K represents the number of sampling points, M is the number of channels of the cosine modulated uniform filter bank, I is the unit column vector, ω p is the passband cutoff frequency of the prototype filter, U, Q, U p , U s is an intermediate variable, and the calculation formula is: U=H(Ω p )U r (Oh p )+H(Ω p -π / M)U r (Oh p -p / M) C(ω)=[cos(N-1)ω / 2,cos(N-3)ω / 2,…,cosω / 2] T In the formula, H(Ω p ), U r (Ω p )、H(Ω p -π / M)、U r (Ω p -π / M)、 Ω p are all intermediate variables, diag is a diagonal matrix, ω s Represents the stopband cutoff frequency of the prototype filter.
4. The method for designing a prototype filter in a cosine modulated non-uniform filter bank according to claim 3, wherein: Step (3) specifically includes: (3.1) Solve the quadratic function model and obtain the optimal solution of the coefficient vector at the tth iteration: in, U q is an intermediate variable; (3.2) According to p (t) Calculate the coefficient vector p for the next iteration (t+1) for: p (t+1) =(1-ξ)p (t) +ξp (t-1) ,t≥1 Where ξ is the smoothing factor, 0<ξ<1; (3.3) Determine whether the following conditions are met. If so, stop the iteration and set p (t) As the optimal solution, and p (t) According to the even symmetry characteristic, the coefficient vector of the second half coefficients is obtained, and then the first half coefficients and the second half coefficients are combined to obtain the coefficient vector of the prototype filter; otherwise, t=t+1 is set and the execution of step (3.1) is returned; ||p (t+1) -p (t) ||2 / ||p (t) ||2<x Where χ represents the error limit.
5. The method for designing a prototype filter in a cosine modulated non-uniform filter bank according to claim 1, characterized in that: 0<α<1,0≤β<1,α+β≤1。 6. A design system for a prototype filter in a cosine modulated non-uniform filter bank, characterized in that: include: The optimization model building module is used to build the prototype filter optimization model: Where g = (g(0), g(1), …, g(N-1)) T is the coefficient vector of the prototype filter, which has an even symmetric characteristic, g(i) is the i-th filter coefficient, N is the number of filter coefficients, α and β are weight factors, E1 is the reconstruction error of the non-uniform filter bank, E2 is the amplitude error of the prototype filter passband, and E3 is the amplitude error of the prototype filter stopband; A model conversion module, used for converting the prototype filter optimization model into a quadratic function model through frequency sampling; The model solving module is used to iteratively solve the quadratic function model and use the coefficient vector at the end of the iteration as the coefficient vector of the optimal prototype filter.
7. The design system of the prototype filter in the cosine modulated non-uniform filter bank according to claim 6, characterized in that: The reconstruction error of the non-uniform filter bank is specifically: Among them, G(e jω )、G(e j(ω-π / M) ) are the frequency responses of the prototype filters of the cosine modulated non-uniform filter bank at frequencies ω and ω-π / M, respectively, and M is the number of channels of the cosine modulated non-uniform filter bank; The amplitude error of the prototype filter passband is specifically: Among them, ω p is the passband cutoff frequency of the prototype filter, ω p =π / 2M; The amplitude error of the prototype filter stop band is specifically: Among them, ω s is the stopband cutoff frequency of the prototype filter, ω s =(π+ε) / 2M, ε is the frequency spacing of adjacent channels in the filter bank.
8. The design system of the prototype filter in the cosine modulated non-uniform filter bank according to claim 6, characterized in that: The quadratic function model is specifically: Where p = (g(0), g(1), ..., g(N / 2-1)) T is the coefficient vector of the first half of the coefficients of the prototype filter, N is the number of filter coefficients, p (t) represents the value of p at the tth iteration, p (0) =0, α, β are weight factors respectively, K represents the number of sampling points, M is the number of channels of the cosine modulated uniform filter bank, I is the unit column vector, ω p is the passband cutoff frequency of the prototype filter, U, Q, U p , U s is an intermediate variable, and the calculation formula is: U=H(Ω p )U r (Oh p )+H(Ω p -π / M)U r (Oh p -p / M) C(ω)=[cos(N-1)ω / 2,cos(N-3)ω / 2,…,cosω / 2] T In the formula, H(Ω p ), U r (Ω p )、H(Ω p -π / M)、U r (Ω p -π / M)、 Ω p are all intermediate variables, diag is a diagonal matrix, ω s Represents the stopband cutoff frequency of the prototype filter.
9. The design system of the prototype filter in the cosine modulated non-uniform filter bank according to claim 8, characterized in that: The model solving module specifically includes: The first solving unit is used to solve the quadratic function model and obtain the optimal solution of the coefficient vector at the t-th iteration: in, U q is an intermediate variable; The second solving unit is used to calculate the value of p (t) Calculate the coefficient vector p for the next iteration (t+1) for: p (t+1) =(1-ξ)p (t) +ξp (t-1) ,t≥1 Where ξ is the smoothing factor, 0<ξ<1; The condition judgment module is used to judge whether the following conditions are met. If they are met, the iteration is stopped and p (t) As the optimal solution, and p (t) According to the even symmetry characteristic, the coefficient vector of the second half coefficients is obtained, and then the first half coefficients and the second half coefficients are combined to obtain the coefficient vector of the prototype filter; otherwise, t=t+1 is set and the first solution unit is returned to execute; ||p (t+1) -p (t) ||2 / ||p (t) ||2<x Where χ represents the error limit.
10. The design system of the prototype filter in the cosine modulated non-uniform filter bank according to claim 6, characterized in that: 0<α<1,0≤β<1,α+β≤1。
