CN111371493B - Machine learning multi-band carrier-free amplitude and phase modulation system based on independent element analysis - Google Patents

Machine learning multi-band carrier-free amplitude and phase modulation system based on independent element analysis Download PDF

Info

Publication number
CN111371493B
CN111371493B CN202010154770.7A CN202010154770A CN111371493B CN 111371493 B CN111371493 B CN 111371493B CN 202010154770 A CN202010154770 A CN 202010154770A CN 111371493 B CN111371493 B CN 111371493B
Authority
CN
China
Prior art keywords
sub
band
signal
interference
ica
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202010154770.7A
Other languages
Chinese (zh)
Other versions
CN111371493A (en
Inventor
迟楠
哈依那尔
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Zhuhai Fudan Innovation Research Institute
Original Assignee
Zhuhai Fudan Innovation Research Institute
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Zhuhai Fudan Innovation Research Institute filed Critical Zhuhai Fudan Innovation Research Institute
Priority to CN202010154770.7A priority Critical patent/CN111371493B/en
Publication of CN111371493A publication Critical patent/CN111371493A/en
Priority to PCT/CN2021/074589 priority patent/WO2021179845A1/en
Application granted granted Critical
Publication of CN111371493B publication Critical patent/CN111371493B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04B—TRANSMISSION
    • H04B10/00—Transmission systems employing electromagnetic waves other than radio-waves, e.g. infrared, visible or ultraviolet light, or employing corpuscular radiation, e.g. quantum communication
    • H04B10/11—Arrangements specific to free-space transmission, i.e. transmission through air or vacuum
    • H04B10/114—Indoor or close-range type systems
    • H04B10/116—Visible light communication
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04B—TRANSMISSION
    • H04B10/00—Transmission systems employing electromagnetic waves other than radio-waves, e.g. infrared, visible or ultraviolet light, or employing corpuscular radiation, e.g. quantum communication
    • H04B10/50—Transmitters
    • H04B10/516—Details of coding or modulation
    • H04B10/548—Phase or frequency modulation

Abstract

The invention belongs to the technical field of visible light communication, and particularly relates to a machine learning multi-band carrierless amplitude phase (multi-band CAP) modulation system based on Independent Component Analysis (ICA). The system mainly comprises modules of matched filtering, down sampling, first-stage CAP forming and cross matched filtering, ICA sub-band signal purification, sub-band signal phase deviation recovery, second-stage CAP forming and cross matched filtering, sub-band interference subtraction, least mean square filtering and the like. According to the invention, after the sub-band interference inversion is carried out at the receiving end, the sub-band interference inversion and the original received sub-band signals are sent to ICA (independent component analysis) to primarily purify each path of sub-band signals, then the sub-band interference inversion of the second stage is carried out, the interference among the sub-bands is subtracted, and the second stage equalization is used, so that the guard interval among the sub-bands can be eliminated, the tolerance degree of the system on the aliasing among the sub-bands is improved, the same transmission rate is realized by using less bandwidth, the signal bandwidth is more effectively utilized, and the frequency spectrum utilization rate of the visible.

Description

Machine learning multi-band carrier-free amplitude and phase modulation system based on independent element analysis
Technical Field
The invention belongs to the technical field of visible light communication, and particularly relates to a machine learning multi-band carrierless amplitude phase modulation (CAP) system based on Independent Component Analysis (ICA) applied to a visible light communication system.
Background
Visible Light Communication (VLC) uses an LED as a Light source, and under the premise of normal lighting, the lighting device can have functions of a wireless router, a Communication base station, a network access point, and the like. The visible light communication as a novel communication mode combining illumination and optical communication will promote the fusion development and technical progress of the next generation of illumination and access networks, and has become the focus and the highest point of competition at home and abroad. The development of visible light communication is of great significance, both at the national strategic level and in its potentially widespread application and large market size.
VLC technology has numerous advantages including no electromagnetic radiation, good safety, high available bandwidth, etc., but its development has also met with some limiting factors, the most significant of which is that the signal modulation bandwidth of commercial white LEDs is very limited, thus limiting the performance of the system. Therefore, it is important to realize a VLC transmission system with high spectral efficiency by fully utilizing a limited bandwidth.
As a high-spectrum-efficiency Amplitude modulation, Carrierless Amplitude and Phase (CAP) has been proved by experiments to be a modulation method very suitable for a visible light communication system, in a multi-sideband CAP modulation system under a multi-user scenario, a certain frequency interval (guard band) must be reserved between subbands, which may cause a low spectrum utilization rate of a VLC system. Meanwhile, the machine learning signal processing method for the communication system is basically based on a data-driven thought, and the communication system is used as a black box to train and learn in a data-driven manner, so that the dependence of a machine learning model on data is greatly increased, and sufficient thinking is not given to the communication system.
Compared with the existing multi-band CAP visible light communication system with reserved frequency band intervals, the invention provides that after sub-band interference inversion is carried out at a receiving end, each path of sub-band signal is preliminarily purified by carrying out Independent Component Analysis (ICA) on the sub-band signal and an original receiving sub-band signal together, then the tolerance degree of the system to inter-sub-band aliasing (overlapping) is improved by carrying out second-stage sub-band interference inversion and subtracting the inter-sub-band interference, and the second-stage equalization is used for realizing the same or higher transmission rate with less bandwidth, so that the signal bandwidth is more effectively utilized, and the spectrum utilization rate of the visible light communication system is improved; and the communication system model is used for processing signals by using a machine learning method, so that the idea of model driving is reflected, the model training efficiency is improved, the data dependence is reduced, and the overall complexity of the system is reduced.
Disclosure of Invention
The invention aims to provide a machine learning multi-band carrierless amplitude-phase modulation (multi-band CAP) system with higher spectrum efficiency based on Independent Component Analysis (ICA); the modulation system is applied to a visible light communication system, can obtain the same or higher system transmission rate by using less modulation bandwidth, and simultaneously, the idea of model driving is reflected in machine learning signal processing, so that the model training efficiency is improved, the data dependence is reduced, and the overall complexity of the system is reduced.
The invention provides a machine learning multi-band carrierless amplitude phase modulation (CAP) system based on Independent Component Analysis (ICA) applied to a visible light communication system, which is composed of modules of matched filtering (& downsampling), first-stage CAP forming and cross matched filtering, ICA sub-band signal purification, sub-band signal recovery phase deviation, second-stage CAP forming and cross matched filtering, sub-band interference subtraction, Least Mean Square (LMS) filtering and the like which are connected in sequence as shown in figure 1; wherein:
the multi-band CAP matched filtering (& down-sampling) module is used for carrying out primary sub-band signal matching separation on the received signals and then obtaining each sub-band complex signal through down-sampling;
the first-stage CAP forming and cross matching filtering module is used for carrying out inversion of interference between first-stage sub-bands and purifying the following ICA sub-band signals;
the ICA subband signal purifying module is used for carrying out preliminary subband signal purification on the signals after the first-stage cross matching filtering and the (interference-carrying) atomic band receiving signals;
the sub-band signal recovery phase deviation module is used for correcting the phase deviation of each sub-band complex signal after the sub-band complex signal passes through the ICA to obtain a preliminary purification signal without the phase deviation;
the second-stage CAP forming and cross matching filtering module is used for carrying out CAP forming and cross matching filtering on each sub-band signal subjected to ICA non-phase-bias preliminary purification to obtain an aliasing interference signal generated by a nearby sub-band in the sub-band;
the sub-band interference subtracting module is used for subtracting the interference of the adjacent sub-bands to the sub-band signals from the ICA-free sub-band signals to obtain relatively pure complex signals;
and the Least Mean Square (LMS) filtering module is used for carrying out second-stage adaptive filtering on each path of sub-band complex signals with interference eliminated to obtain signals with less interference for further error rate testing.
In the invention, at the transmitting end, each path of sub-band signal accords with the general processing process of the signal in the traditional CAP modulation system, and finally, all the modulated sub-band CAP signals are added and transmitted, which is not explained more here. In the embodiment simulation, the modulation order of the system design is as follows: 16 QAM. In the simulation example, two sub-band CAP signals are used for illustration, and the method can be extended to three sub-band signals and multi-sub-band CAP signals without loss of generality.
In the present invention, the multi-band CAP matched filtering (&Downsampling) module for performing preliminary subband signal matched filtering separation on the received signal, and then downsampling to obtain each subband complex signal, wherein the nth subband signal is represented as rn(t)＝rni(t)+i×rnq(t), wherein the real part and the imaginary part of each signal are respectively obtained by the following equations:
Figure BDA0002403688380000031
wherein m isnI(t)＝fnI(-t)，where fnI(t)＝g(t)*cos(2πfcnt) is the time domain response of the n path of I path matched filter, mnQ(t)＝fnQ(-t)，where fnQ(t)＝g(t)*sin(2πfcnt) is the time domain response of the nth path of Q path matched filter; and then, using each path of sub-band complex signal obtained for the first-stage CAP forming and cross matching filtering to invert the interference.
In the invention, the first stage CAP forming and cross matching filtering module is used for obtaining aliasing interference between sub-bands, wherein CAP signal forming is to filter an I path and a Q path of each sub-band complex signal by using a pair of forming filters satisfying the characteristic of hilbert transform pair, and the up-sampled sub-band forming filtering satisfies the following relational expression:
Figure BDA0002403688380000032
wherein s isn(t) is the nth sub-band CAP signal, an(t) and bn(t) represents the I path component and the Q path component of the n path of the up-sampled signal respectively, fnI(t)＝g(t)*cos(2πfcnt)、fnQ(t)＝g(t)*sin(2πfcnt) respectively representing the time domain response of the I path of homodromous shaping filter and the time domain response of the Q path of orthogonal shaping filter in the nth path of subband complex signal, wherein g (t) is a baseband square root raised cosine Nyquist filter, fcnIs the nth carrier frequency; IQ two paths of signals after each sub-band signal is filtered are in a quadrature relationship, and each path of CAP signal is regenerated through a subtraction operation; in the following cross matching filtering, after normalization processing is performed on each channel of sub-band CAP signals after reshaping, cross matching filtering is used to obtain interference signals from other sub-bands nearby in each sub-band, and interference from a second sub-band (in the middle frequency band) in a first sub-band (in the lower frequency band) is represented as:
ISI12(t)＝ISI12i(t)+i×ISI12q(t) (3)
wherein the content of the first and second substances,
Figure BDA0002403688380000033
IQ two-way filtering is carried out on the complex signal of the first sub-band by using a pair of matched filters of the second sub-band to obtain aliasing interference ISI generated by the second sub-band in the first sub-band12(t); similarly, the interference in the second subband (in the middle band) from the first subband (in the lower band) and the third subband (in the upper band) can be expressed as:
ISI213(t)＝ISI213i(t)+i×ISI213q(t) (4)
wherein the content of the first and second substances,
Figure BDA0002403688380000041
i.e. the interference in the second subband can be divided into two parts, one from the first subband (in the lower band) in the second subband ((R))In the middle band) of the interference ISI21(t)＝ISI21i(t)+i×ISI21q(t), interference ISI from a third subband (in the higher frequency band) in a second subband (in the middle frequency band)23(t)＝ISI23i(t)+i×ISI23q(t); similarly, the interference from the second subband (in the middle band) in the third subband (in the higher band) is represented as:
ISI32(t)＝ISI32i(t)+i×ISI32q(t) (6)
wherein the content of the first and second substances,
Figure BDA0002403688380000042
using a pair of matched filters of the second sub-band (in the middle frequency band) to carry out IQ two-path filtering on the complex signal of the third sub-band (in the higher frequency band) to obtain the aliasing interference ISI generated by the second band in the third band32(t); wherein, the time domain response of each matched filter is the same as the description in the claim 2, which is not repeated herein; and then, sending each sub-band signal and the inverted corresponding in-band interference signal into an ICA module to purify each sub-band signal.
In the ICA subband signal refining module, each subband signal and the inverted corresponding inband interference signal are sent to the ICA, for example, the first subband signal band1 and the ISI signal from the second band interfering with the first subband signal band1 are sent to the ICA12(t) ICA is carried out together, and a first sub-band signal band1 (in a lower frequency band) is preliminarily separated; the same applies to the processing of the other subbands, namely the second subband band2 and the interference signal ISI from the first and third subbands21(t)、ISI23(t) ICA together, preliminarily separating a second subband signal band2 (in the middle band); combining the third sub-band 3 with the interference signal ISI from the second sub-band32(t) together with the ICA, a third subband signal band3 is initially isolated (in the higher frequency band). The matrix R formed by the sub-band complex signals of each path is sent to the ICA; the ICA process is as follows:
(1) r mean removal whitening, the whitening matrix is:
Figure BDA0002403688380000043
obtaining a whitened sample
Figure BDA0002403688380000044
Wherein λn、
Figure BDA0002403688380000045
Respectively a characteristic root and a characteristic vector of R;
(2) randomly initializing and normalizing weight matrices
Figure BDA0002403688380000046
(3) Calculating a cumulative distribution function:
Figure BDA0002403688380000047
calculating and normalizing the weight w;
(4) weight iterative formula:
Figure BDA0002403688380000051
iterate until the weight w converges.
After ICA, matrix S composed of preliminarily purified subband complex signals and interference signals is obtained, but the front and back sequence of the signals is disturbed, namely the corresponding positions of the R matrix and the S matrix are not necessarily signals corresponding to the front and back of ICA, and each subband signal generates phase deviation in different degrees, so that subband identification needs to be carried out on the S matrix, and then the purified signals are sent to a next subband signal recovery phase deviation module.
In the invention, each sub-band signal restores a phase deviation module, a unique known training sequence tx is respectively added to each sub-band signal of an R matrix in the module, and the phase information is expressed as
Figure BDA0002403688380000052
This training sequence after ICA is denoted as rx, and its phase information is denoted as rx
Figure BDA0002403688380000053
Thus the angle e of phase deviation before and after ICAiΔθDividing the received training sequence by the transmitted training sequence, and then averaging to obtain:
Figure BDA0002403688380000054
then multiplying each path of sub-band signals after ICA by ei(-Δθ)The phase deviation angle can be recovered; after each sub-band signal is identified and the phase deviation is corrected, each sub-band signal which is recovered by ICA post-phase deviation can be obtained, and is a plurality of signals of each path of preliminary purification, and then the multi-path signals are sent to a second-stage CAP forming and cross matching filter module.
In the present invention, the operation of the second stage CAP forming and cross-matched filtering module is the same as that of the first stage CAP forming and cross-matched filtering module, which is not described herein again. And after the inversion interference signal of the second stage is obtained, a module for subtracting the sub-band interference (obtaining a pure signal) from the sub-band signal of the primary purification is carried out in the next step.
In the present invention, the sub-band interference module is subtracted from the ICA preliminary purification sub-band signal, that is, each interference signal obtained in claim 6 is subtracted from the preliminary purification signal (after sub-band identification and phase offset angle recovery) after ICA, which can be expressed as follows:
Figure BDA0002403688380000055
wherein s ispure(t) represents the cleaner subband signals after the subband interference is subtracted, and weight represents the weight of the subtracted interference; then sends the purer signal to the next stepA Least Mean Square (LMS) filtering module.
In the invention, the Least Mean Square (LMS) filtering module is used for each path of sub-band complex signal s with interference eliminatedpure(t) performing second-stage LMS adaptive filtering to obtain a cleaner signal; and finally, carrying out error rate test.
In an embodiment, the filter tap coefficient is 27, the step size is 0.004, and the training sequence length is 2000;
compared with the existing multi-band CAP visible light communication system with reserved frequency band intervals, the invention provides that at a receiving end, after sub-band interference inversion, the sub-band interference inversion and the original received sub-band signals are sent to ICA to primarily purify each path of sub-band signals, then the sub-band interference inversion of the second stage is carried out, the interference among the sub-bands is subtracted, and the second stage of equalization is used to improve the tolerance degree of the system to the inter-sub-band aliasing (overlapping), the same or higher transmission rate can be realized by using less bandwidth, the signal bandwidth is more effectively utilized, and the frequency spectrum utilization rate of the visible light communication system is improved; and the communication system model is used for processing signals by using a machine learning method, so that the idea of model driving is reflected, the model training efficiency is improved, the data dependence is reduced, and the overall complexity of the system is reduced.
It can be seen from the modulation system provided by the present invention that the method using machine learning driven by model through inverting subband interference and eliminating the subband interference has the following advantages:
(1) the model-driven ICA machine learning receiving method is used for multi-band carrierless amplitude phase modulation CAP for the first time, and the same or higher system transmission rate can be obtained by using less modulation bandwidth by greatly improving the tolerance of the degree of mutual aliasing (overlapping) among sub-bands;
(2) after each path of sub-band signals are primarily purified and separated through the first-level ICA, sub-band interference is inverted and aliasing interference among sub-bands is eliminated, so that not only can the frequency interval between the traditional multi-band CAP sub-bands be eliminated, but also the tolerance of the degree of aliasing among the sub-bands is greatly improved, and the system capacity is further improved;
(3) the ICA-based machine learning multi-band CAP receiving method for inverting and eliminating the inter-subband interference is applied to a visible light communication system for the first time, simulation is carried out, and the visible light communication system with higher spectral efficiency is proved by theory;
(4) the method is provided for processing signals by using a machine learning method in a visible light communication system from a communication system model, so that the idea of model driving is embodied, the model training efficiency is improved, the data dependence is reduced, and the overall complexity of the system is reduced.
The invention is suitable for the field of short-distance, high-speed and high-spectrum-efficiency visible light communication, provides a multi-user scene solution with compressed bandwidth, and provides a machine learning method for processing signals from a communication system model, so that the idea of model driving is embodied.
Drawings
Fig. 1 is a diagram of an ICA-based machine learning multi-band CAP receiving system of the present invention.
Fig. 2 is a schematic diagram of inter-subband spectral aliasing in the multi-band CAP receiving scheme of the present invention (the number of subbands is 2 in the example, and the subband aliasing ratio is 42%).
Fig. 3 is a study of error performance variation with snr in a multi-band CAP16 system according to the present invention.
Fig. 4 is a study of error performance variation with inter-subband aliasing ratio in a multi-band CAP16 system by the ICA-based machine learning multi-band CAP receiving method of the present invention.
Reference numbers in the figures: 101 is a matched filtering & down sampling module, 102 is a first-stage CAP forming and cross matched filtering module, 103 is an ICA sub-band signal purifying module, 104 is a phase deviation restoring module for each sub-band signal, 105 is a second-stage CAP forming and cross matched filtering module, 106 is a sub-band interference subtracting module, and 107 is a least mean square filtering module.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, embodiments of the present invention will be described in detail below with reference to the accompanying drawings. However, it will be appreciated by those of ordinary skill in the art that numerous technical details are set forth in order to provide a better understanding of the present application in various embodiments of the present invention. However, the technical solutions claimed in the claims of the present application can be implemented without these technical details and with various changes and modifications based on the following embodiments.
The modulation system provided by the invention is applied to a visible light communication system, can obtain the same system transmission rate by using less modulation bandwidth, and obtains higher frequency spectrum utilization rate; the idea and method for processing the machine learning signal driven by the model in the visible light communication system are also provided. The model training efficiency is improved, the data dependence is reduced, and the overall complexity of the system is reduced.
The modulation system proposed by the present invention is shown in fig. 1, and comprises matched filtering (& downsampling), first stage CAP forming and cross matched filtering, ICA subband signal purification, each subband signal phase deviation recovery, second stage CAP forming and cross matched filtering,
Subtracting sub-band interference, Least Mean Square (LMS) filtering, etc.
The machine learning multi-band carrierless amplitude phase modulation (CAP) receiving method based on the Independent Component Analysis (ICA) of the modulation system comprises the following specific steps.
Step 101: performing preliminary sub-band signal matching separation on the received signal through a multi-band CAP matching filtering (& down sampling) module, and then performing down sampling to obtain each sub-band complex signal;
this is a general matched filtering process of the received signal in the traditional multi-band CAP modulation system, and we do not describe it too much, and the modulation order of each sub-band signal designed by this system is: 16 QAM; two subbands are exemplified in this example. And then, using each path of sub-band complex signal obtained for the first-stage CAP forming and cross matching filtering to invert the interference.
Step 102: performing primary CAP forming and cross matching filtering on each subband signal obtained by primary matching filtering through a primary CAP forming and cross matching filtering module to realize the inversion of the interference between the first-stage subbands;
performing first-stage CAP forming and cross matching filtering on each path of subband signals obtained in step 101 to obtain aliasing interference between subbands, wherein the CAP signal forming is to filter an I path and a Q path of each subband complex signal by using a pair of forming filters satisfying the characteristic of hilbert transform pair, and the upsampled each path of subband forming filtering satisfies the following relational expression:
Figure BDA0002403688380000071
wherein s isn(t) is the nth sub-band CAP signal, an(t) and bn(t) represents the I path component and the Q path component of the n path of the up-sampled signal respectively, fnI(t)＝g(t)*cos(2πfcnt)、fnQ(t)＝g(t)*sin(2πfcnt) respectively representing the time domain response of the I path of homodromous shaping filter and the time domain response of the Q path of orthogonal shaping filter in the nth path of subband complex signal, wherein g (t) is a baseband square root raised cosine Nyquist filter, fcnIs the nth carrier frequency; IQ two paths of signals after each sub-band signal is filtered are in a quadrature relationship, and each path of CAP signal is regenerated through a subtraction operation; in the following cross matching filtering, after normalization processing is performed on each channel of sub-band CAP signals after reshaping, cross matching filtering is used to obtain interference signals from other sub-bands nearby in each sub-band, and interference from a second sub-band (in the middle frequency band) in a first sub-band (in the lower frequency band) is represented as:
ISI12(t)＝ISI12i(t)+i×ISI12q(t)
wherein the content of the first and second substances,
Figure BDA0002403688380000081
IQ two-way filtering is carried out on the complex signal of the first sub-band by using a pair of matched filters of the second sub-band to obtain aliasing interference ISI generated by the second sub-band in the first sub-band12(t); similarly, the second sub-band (at the intermediate frequency)Segment) and the interference from the first subband (in the lower band) and the third subband (in the higher band) can be expressed as:
ISI213(t)＝ISI213i(t)+i×ISI213q(t)
wherein the content of the first and second substances,
Figure BDA0002403688380000082
i.e. the interference in the second subband can be divided into two parts, one is the interference ISI from the first subband (in the lower frequency band) generated in the second subband (in the middle frequency band)21(t)＝ISI21i(t)+i×ISI21q(t), interference ISI from a third subband (in the higher frequency band) in a second subband (in the middle frequency band)23(t)＝ISI23i(t)+i×ISI23q(t); similarly, the interference from the second subband (in the middle band) in the third subband (in the higher band) is represented as:
ISI32(t)＝ISI32i(t)+i×ISI32q(t)
wherein the content of the first and second substances,
Figure BDA0002403688380000083
using a pair of matched filters of the second sub-band (in the middle frequency band) to carry out IQ two-path filtering on the complex signal of the third sub-band (in the higher frequency band) to obtain the aliasing interference ISI generated by the second band in the third band32(t); wherein, the time domain response of each matched filter is the same as the description in the claim 2, which is not repeated herein; and then, sending each sub-band signal and the inverted corresponding in-band interference signal into an ICA module to purify each sub-band signal.
Step 103: carrying out ICA on the interference signal after the first-stage cross matching filtering and the original sub-band receiving signal (with interference) by an ICA sub-band signal purification module to realize preliminary sub-band signal purification and separation;
sending each sub-band signal obtained in step 102 and the inverted corresponding in-band interference signal into ICA, for example, sending the first sub-band signal band1 and the signal from the second band pairIts interfering signal ISI12(t) ICA is carried out together, and a first sub-band signal band1 (in a lower frequency band) is preliminarily separated; the same applies to the processing of the other subbands, namely the second subband band2 and the interference signal ISI from the first and third subbands21(t)、ISI23(t) ICA together, preliminarily separating a second subband signal band2 (in the middle band); combining the third sub-band 3 with the interference signal ISI from the second sub-band32(t) together with the ICA, a third subband signal band3 is initially isolated (in the higher frequency band). ICA procedure is omitted. After ICA, matrix S composed of preliminarily purified subband complex signals and interference signals is obtained, but the front and back sequence of the signals is disturbed, namely the corresponding positions of the R matrix and the S matrix are not necessarily signals corresponding to the front and back of ICA, and each subband signal generates phase deviation in different degrees, so that subband identification needs to be carried out on the S matrix, and then the purified signals are sent to a next subband signal recovery phase deviation module.
Step 104: correcting the phase deviation of each sub-band complex signal after ICA through a sub-band signal recovery phase deviation module to obtain a preliminary purification separation signal without phase deviation;
in this step, a unique known training sequence tx is added to each path of sub-band signal of R matrix to be ICA, and its phase information is expressed as
Figure BDA0002403688380000091
This training sequence after ICA is denoted as rx, and its phase information is denoted as rx
Figure BDA0002403688380000092
Thus the angle e of phase deviation before and after ICAiΔθDividing the received training sequence by the transmitted training sequence, and then averaging to obtain:
Figure BDA0002403688380000093
then multiplying each path of sub-band signals after ICA by ei(-Δθ)The phase deviation angle can be recovered; after each sub-band signal is identified and the phase deviation is corrected, each sub-band signal which is recovered by ICA post-phase deviation can be obtained, and is a plurality of signals of each path of preliminary purification, and then the multi-path signals are sent to a second-stage CAP forming and cross matching filter module.
Step 105: performing second-stage CAP forming and cross matching filtering on each sub-band signal subjected to ICA non-phase-bias preliminary purification through a second-stage CAP forming and cross matching filtering module to obtain an aliasing interference signal generated by a nearby sub-band in the sub-band;
the operation of this step is the same as that described in step 102, and is not described here. And after the inversion interference signal of the second stage is obtained, a module for subtracting the sub-band interference (obtaining a pure signal) from the sub-band signal of the primary purification is carried out in the next step.
Step 106: by subtracting sub-band interference, subtracting the interference of adjacent sub-bands to each sub-band signal after ICA to obtain a relatively pure complex signal; adjusting the weight coefficient to obtain the best interference elimination effect;
subtracting each interference signal found in step 105 from the preliminary refined signal (after subband identification and phase offset angle recovery) after ICA can be expressed as follows:
s1pure(t)＝s1(t)-weight×ISI12(t)
s2pure(t)＝s2(t)-weight×ISI21(t)-weight×ISI23(t)
s3pure(t)＝s3(t)-weight×ISI32(t)
wherein s ispure(t) represents the cleaner respective subband signals after subtracting the subband interference, and weight represents the weight of the subtracted interference.
Step 107: and performing second-stage Least Mean Square (LMS) adaptive filtering on each sub-band complex signal with interference eliminated through modules such as LMS filtering and the like to obtain a cleaner signal for further error rate testing.
Carrying out second-stage adaptive filtering on each path of sub-band complex signals with interference eliminated to obtain each sub-band signal with cleaner star; wherein the tap coefficient of the filter is 27, the step length is 0.004, and the length of the training sequence is 2000; and finally, carrying out error rate test.
In the present invention, the parameters of the FastICA algorithm used by ICA can be adjusted during the simulation to obtain the best separation effect. The LMS tap number and step size can also be adjusted appropriately to achieve the best equalization effect.
So far, the ICA-based machine learning multiband CAP reception method ends.
The verification steps and the verification results of the ICA-based machine learning multiband CAP receiving method in the visible light communication simulation system are described next. The method comprises the following specific steps:
in this example, the VLC system simulation framework can also be obtained in fig. 1, where in fig. 1, reference numeral 101 denotes a matched filtering & downsampling module, reference numeral 102 denotes a first-stage CAP forming and cross-matched filtering module, reference numeral 103 denotes an ICA subband signal purifying module, reference numeral 104 denotes each subband signal phase deviation restoring module, reference numeral 105 denotes a second-stage CAP forming and cross-matched filtering module, reference numeral 106 denotes a subband interference subtracting module, and reference numeral 107 denotes a least mean square filtering module. In the example, an exponential function is used for simulating a high-frequency fading effect H of a visible light channel; second, Additive White Gaussian Noise (AWGN) with a signal-to-noise ratio of SNR is introduced. And then, matching and filtering at a receiving end to obtain each path of sub-band signals of the multi-band CAP signal.
Fig. 2 is a schematic diagram of an inter-subband spectral aliasing region in a multi-band CAP16 visible light communication system. Wherein red is the spectral envelope of the CAP signal of the first sub-band and blue is the spectral envelope of the CAP signal of the second sub-band; the region numbered 21 is a region where subbands are aliased with each other, and here we can see that the ICA-based machine learning multiband CAP receiving method provided by the invention can not only eliminate guard bands (guard bands) between subbands in the traditional multiband CAP VLC system, but also can tolerate approximately 42% of subband aliasing with each other, thereby improving the utilization efficiency of frequency spectrum.
Fig. 3 is a graph for studying the change of the error code performance along with the signal-to-noise ratio in a multi-band CAP16 VLC simulation system by the ICA-based machine learning multi-band CAP receiving method provided by the invention. The numbers 31 and 32 respectively represent the curves of the error performance of the first sub-band signal band1 and the second sub-band signal band2 after the first stage of ICA purification separation along with the signal-to-noise ratio, it can be seen that the BER of both sub-bands decreases and the error performance increases along with the increase of the signal-to-noise ratio SNR, and the two curves are almost overlapped, which indicates that the error performance of both sub-bands after the ICA purification is basically consistent, which is also related to that both sub-bands are basically in the flat region of the visible light channel, so that the signal in the higher frequency band is not faded too much. Numbers 33 and 34 respectively represent curves of error code performance of the first sub-band signal band1 and the second sub-band signal band2 after sub-band interference inversion and cancellation along with signal-to-noise ratio changes, and it can be seen that after sub-band interference inversion and cancellation, BER is reduced by orders of magnitude, and error code performance is greatly improved. And finally, when the signal-to-noise ratio is high, the BER of the two sub-bands is reduced below the error code threshold. The error rate is the error rate of 3.8e-3 considering the FEC threshold of 7%, which is also referred to in the following, and is not described again.
Fig. 4 is a graph for studying the change of the error code performance along with the aliasing ratio among sub-bands in a multi-band CAP16 VLC simulation system by the ICA-based machine learning multi-band CAP receiving method provided by the invention. The number 41, the number 42 and the number 43 respectively represent curves of the error performance of the first subband band1 after no equalization processing, the first-stage ICA purification separation, the subband interference inversion and cancellation and the second-stage equalization (LMS) along with the aliasing ratio among the subbands. It can be seen that the BER of the sub-band signals decreases sequentially with the two-stage processing, and when the aliasing proportion between the sub-bands reaches 42%, the transmission error rate of the sub-band signals can still be ensured to be lower than the FEC error threshold of 7%. As the aliasing ratio increases, BER increases and error performance decreases. Meanwhile, we can also see that when aliasing is not serious (for example, aliasing seen in a simulation result is less than 30%), the problem can be solved by the first-level ICA purification subband signal; when aliasing is serious (for example, aliasing seen in a simulation result is greater than 36%), two-stage processing (ICA proposed & inverse interference cancellation) can further improve the degree of aliasing of the spectrum between the sub-bands and improve the spectrum utilization rate. Reference numeral 44, reference numeral 45 and reference numeral 46 respectively show the curves of the error performance of the second subband band2 after no equalization, the first stage ICA refinement and separation, the subband interference inversion and cancellation, and the second stage equalization (LMS) as a function of the aliasing ratio between the subbands. The analysis result is the same as that of band1, and is not described in detail here.
More intuitively, the lower side of fig. 4 shows the corresponding constellation diagram of two subband signals after no equalization processing, the first level ICA purification and separation, the subband interference inversion and elimination, and the second level equalization (LMS) when the aliasing ratio between the subbands reaches 42%. A. C, E are the corresponding constellation diagrams of the two sub-band signals after the first sub-band signal band1 is not processed by any equalization, is processed by the first level ICA purification and separation, is processed by the sub-band interference inversion and elimination and is processed by the second level equalization (LMS); B. d, F are the corresponding constellation diagrams of the two sub-band signals after the second sub-band signal band2 is not processed by any equalization, is processed by the first level ICA purification and separation, is processed by the sub-band interference inversion and elimination and is processed by the second level equalization (LMS); it can be seen that after ICA and sub-band interference of the original received sub-band signal are inverted and eliminated, the constellation diagram becomes clearer and better in error code performance.
In summary, through simulation comparison of multiple dimensions, the present embodiment shows, as a simulation result, that compared to the existing multiband CAP receiving method, the proposed ICA-based machine learning multiband CAP receiving method has a higher spectrum utilization rate in the multiband CAP VLC system, and can achieve the same or higher transmission rate with less bandwidth. The performance of the traditional multi-band CAP visible light communication system is greatly improved due to the ICA-based machine learning multi-band CAP receiving method provided by the invention. Compared with the traditional data-driven machine learning signal processing method, the ICA-based machine learning multiband CAP receiving method provided by the invention uses the machine learning method to process signals from the communication system model, so that the idea of model driving is reflected, the model training efficiency is improved, the data dependence is reduced, and the overall complexity of the system is reduced.
The division of each step in this embodiment is only for clarity of description, and implementation may be combined into one step or split some steps into multiple steps, and all that is included in the same logical relationship is within the scope of the present patent.
It will be understood by those of ordinary skill in the art that the foregoing embodiments are specific examples for carrying out the invention, and that various changes in form and details may be made therein without departing from the spirit and scope of the invention in practice.

Claims (5)

1. A machine learning multi-band carrier-free amplitude and phase modulation system based on independent element analysis is characterized by being formed by sequentially connecting a multi-band CAP matched filtering & down-sampling module, a first-stage CAP forming and cross matched filtering module, an ICA sub-band signal purification module, sub-band signal recovery phase deviation modules, a second-stage CAP forming and cross matched filtering module, a sub-band interference subtraction module and a least mean square LMS filtering module; wherein:
the multi-band CAP matched filtering and down-sampling module is used for carrying out primary sub-band signal matched separation on the received signals and then obtaining each sub-band complex signal through down-sampling;
the first-stage CAP forming and cross matching filtering module is used for carrying out inversion of interference between first-stage sub-bands and purifying the following ICA sub-band signals;
the ICA subband signal purifying module is used for carrying out preliminary subband signal purification on the signals after the first-stage cross matching filtering and the atomic band receiving signals;
the sub-band signal recovery phase deviation module is used for correcting the phase deviation of the sub-band complex signals after the ICA sub-band signal purification module to obtain a preliminary purification signal without phase deviation;
the second-stage CAP forming and cross matching filtering module is used for performing CAP forming and cross matching filtering on preliminarily purified subband signals without phase deviation obtained after the ICA subband signal purifying module and the subband signal restoring phase deviation module, and further obtaining residual aliasing interference signals generated by adjacent subbands in the subbands;
the sub-band interference subtracting module is used for subtracting the interference of the adjacent sub-bands to the sub-band signals from the ICA-free sub-band signals to obtain relatively pure complex signals;
the least mean square LMS filtering module is used for carrying out second-stage adaptive filtering on each path of sub-band complex signals with interference eliminated to obtain signals with less interference for further error rate test;
the multi-band CAP matched filtering&The down-sampling module is used for carrying out primary sub-band signal matching filtering separation on the received signals and then obtaining each sub-band complex signal through down-sampling; wherein, let the nth sub-band signal be denoted as rn(t)＝rni(t)+i×rnq(t), the real part and the imaginary part of each signal are respectively obtained by the following equations:
Figure FDA0003006703720000011
wherein m isnI(t)＝fnI(-t)，where fnI(t)＝g(t)*cos(2πfcnt) is the time domain response of the n path of I path matched filter, mnQ(t)＝fnQ(-t)，where fnQ(t)＝g(t)*sin(2πfcnt) is the time domain response of the nth path of Q-path matched filter, fnI(t)＝g(t)*cos(2πfcnt)、fnQ(t)＝g(t)*sin(2πfcnt) respectively representing the time domain response of the I path of homodromous shaping filter and the time domain response of the Q path of orthogonal shaping filter in the nth path of subband complex signal, wherein g (t) is a baseband square root raised cosine Nyquist filter, fcnIs the nth carrier frequency;
Figure FDA0003006703720000021
represents a convolution operation, the same below; then, each path of sub-band complex signal is used for the first stage CAP forming and cross matching filtering to invert the trunkDisturbing;
the first-stage CAP forming and cross-matching filtering module is used for solving aliasing interference among subbands, wherein CAP signal forming is to filter an I path and a Q path of each subband complex signal by using a pair of forming filters meeting the characteristic of a Hilbert transform pair, and each subband forming filter after upsampling meets the following relational expression:
Figure FDA0003006703720000022
wherein s isn(t) is the nth sub-band CAP signal, an(t) and bn(t) represents the I path component and the Q path component of the n path of the up-sampled signal respectively, fnI(t)＝g(t)*cos(2πfcnt)、fnQ(t)＝g(t)*sin(2πfcnt) respectively representing the time domain response of the I path of homodromous shaping filter and the time domain response of the Q path of orthogonal shaping filter in the nth path of subband complex signal, wherein g (t) is a baseband square root raised cosine Nyquist filter, fcnIs the nth carrier frequency; IQ two paths of signals after each sub-band signal is filtered are in a quadrature relationship, and each path of CAP signal is regenerated through a subtraction operation; in the following cross matching filtering, normalization processing is performed on each channel of sub-band CAP signals after reshaping, then cross matching filtering is used to obtain interference signals from other sub-bands nearby in each sub-band, and interference from a second sub-band in a first sub-band is represented as:
ISI12(t)＝ISI12i(t)+i×ISI12q(t) (3)
wherein the content of the first and second substances,
Figure FDA0003006703720000023
IQ two-way filtering is carried out on the complex signal of the first sub-band by using a pair of matched filters of the second sub-band to obtain aliasing interference ISI generated by the second sub-band in the first sub-band12(t); similarly, the interference in the second subband from the first subband and the third subband is representedComprises the following steps:
ISI213(t)＝ISI213i(t)+i×ISI213q(t) (4)
wherein the content of the first and second substances,
Figure FDA0003006703720000024
i.e. the interference in the second subband is divided into two parts:
one is the interference from the first subband in the second subband: ISI (inter-symbol interference)21(t)＝ISI21i(t)+i×ISI21q(t)，
Second is interference from the third subband in the second subband: ISI (inter-symbol interference)23(t)＝ISI23i(t)+i×ISI23q(t)；
Similarly, the interference from the second subband in the third subband is represented as:
ISI32(t)＝ISI32i(t)+i×ISI32q(t) (6)
wherein the content of the first and second substances,
Figure FDA0003006703720000031
IQ two-path filtering is carried out on the complex signal of the third sub-band by using a pair of matched filters of the second sub-band to obtain aliasing interference ISI generated by the second sub-band in the third sub-band32(t); then, each sub-band signal and the corresponding inverted in-band interference signal are sent to an ICA module for purification of each sub-band signal;
the ICA sub-band signal purification module is used for sending each sub-band signal and the corresponding in-band interference signal obtained by inversion into the ICA; the ICA is fed with a matrix R consisting of subband complex signals and corresponding inband interfering signals, and then the ICA is processed as follows:
(1) r mean removal whitening, the whitening matrix is:
Figure FDA0003006703720000032
obtaining a whitened sample
Figure FDA0003006703720000033
Wherein λn、
Figure FDA0003006703720000034
Respectively a characteristic root and a characteristic vector of R;
(2) randomly initializing and normalizing weight matrices
Figure FDA0003006703720000035
(3) Calculating a cumulative distribution function:
Figure FDA0003006703720000036
calculating and normalizing the weight w;
(4) weight iterative formula:
Figure FDA0003006703720000037
iterating until the weight w converges;
after ICA, obtaining a matrix S formed by the preliminarily purified sub-band complex signals of each path and the interference signals; performing sub-band identification on the S matrix, and then sending the purified signal to a next sub-band signal recovery phase deviation module;
and the sub-band signals recover the phase deviation module, wherein, a section of unique known training sequence tx is respectively added to each path of sub-band signals of the R matrix, and the phase information is expressed as
Figure FDA0003006703720000038
This training sequence after ICA is denoted as rx, and its phase information is denoted as rx
Figure FDA0003006703720000039
Such that the front and rear phase angles of the ICAeiΔθDividing the received training sequence by the transmitted training sequence, and then averaging to obtain:
Figure FDA00030067037200000310
then multiplying each path of sub-band signals after ICA by ei(-Δθ)The phase deviation angle can be recovered; after each sub-band signal is identified and the phase deviation is corrected, each sub-band signal which is recovered by the phase deviation after ICA is obtained, and is a complex signal of each primary purification, and then the multi-channel signal is sent to a second-stage CAP forming and cross matching filter module.
2. The system according to claim 1, wherein the second stage CAP shaping and cross-matched filtering module has the same working contents as the first stage CAP shaping and cross-matched filtering module, but its input is the first round of preliminary purified complex signals of each sub-band, and its output is the residual interference signals from the nearby sub-bands obtained by inversion, and the residual sub-band interference is subtracted from the preliminary purified sub-band signals.
3. The system of claim 2, wherein the sub-band interference subtraction module subtracts interference of neighboring sub-bands from each sub-band signal after ICA without phase offset to obtain a relatively pure complex signal; the method specifically comprises the following steps:
subtracting the inverted interference signal of the second stage from the preliminary refined signal after ICA, as follows:
Figure FDA0003006703720000041
wherein s ispure(t) indicates the purer ones after subtraction of the sub-band interferenceSubband signals, weight represents the weight of the interference being subtracted; the cleaner signal is then sent to the next least mean square filtering module.
4. The system according to claim 3, wherein the MMSE-based machine learning multi-band carrierless amplitude phase modulation system is configured to apply the least mean square filtering module to each sub-band complex signal s with interference eliminatedpure(t) performing second-stage LMS adaptive filtering to obtain a cleaner signal; wherein the tap coefficient of the filter is 27, the step length is 0.004, and the length of the training sequence is 2000; and finally, carrying out error rate test.
5. An independent component analysis-based machine learning multi-band carrierless amplitude and phase modulation method based on the system of any one of claims 1 to 4, characterized by comprising the following specific steps:
step 101: performing preliminary sub-band signal matching separation on the received signals through a multi-band CAP matching filtering and down-sampling module, and then performing down-sampling to obtain each sub-band complex signal;
step 102: performing primary CAP forming and cross matching filtering on each subband complex signal obtained by primary matching filtering through a primary CAP forming and cross matching filtering module to realize the inversion of the interference between the first-stage subbands;
step 103: carrying out ICA on the signal subjected to the first-level cross matching filtering and the original sub-band receiving signal by an ICA sub-band signal purification module so as to realize primary sub-band signal purification and separation;
step 104: correcting the phase deviation of each sub-band complex signal after ICA through a sub-band signal recovery phase deviation module to obtain a preliminary purification separation signal without phase deviation;
step 105: performing second-stage CAP forming and cross matching filtering on each sub-band signal subjected to ICA non-phase-bias preliminary purification through a second-stage CAP forming and cross matching filtering module to obtain an aliasing interference signal generated by a nearby sub-band in the sub-band;
step 106: through a sub-band interference subtraction module, subtracting the interference of a nearby sub-band to each sub-band signal after ICA from each sub-band signal without phase deviation to obtain a relatively pure complex signal; adjusting the weight coefficient to obtain the best interference elimination effect;
step 107: and performing second-stage least mean square adaptive filtering on each path of sub-band complex signals with interference eliminated through a least mean square filtering module to obtain cleaner signals for further error rate testing.
CN202010154770.7A 2020-03-08 2020-03-08 Machine learning multi-band carrier-free amplitude and phase modulation system based on independent element analysis Active CN111371493B (en)

Priority Applications (2)

Application Number Priority Date Filing Date Title
CN202010154770.7A CN111371493B (en) 2020-03-08 2020-03-08 Machine learning multi-band carrier-free amplitude and phase modulation system based on independent element analysis
PCT/CN2021/074589 WO2021179845A1 (en) 2020-03-08 2021-02-01 Machine learning multi-band carrierless amplitude and phase modulation system based on independent component analysis

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010154770.7A CN111371493B (en) 2020-03-08 2020-03-08 Machine learning multi-band carrier-free amplitude and phase modulation system based on independent element analysis

Publications (2)

Publication Number Publication Date
CN111371493A CN111371493A (en) 2020-07-03
CN111371493B true CN111371493B (en) 2021-06-04

Family

ID=71211184

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010154770.7A Active CN111371493B (en) 2020-03-08 2020-03-08 Machine learning multi-band carrier-free amplitude and phase modulation system based on independent element analysis

Country Status (2)

Country Link
CN (1) CN111371493B (en)
WO (1) WO2021179845A1 (en)

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111371493B (en) * 2020-03-08 2021-06-04 珠海复旦创新研究院 Machine learning multi-band carrier-free amplitude and phase modulation system based on independent element analysis
CN114422028A (en) * 2021-12-29 2022-04-29 中国电信股份有限公司 Signal demodulation method and device, electronic equipment and readable storage medium
CN114726405B (en) * 2022-03-09 2023-06-27 南京邮电大学 Power line communication method, device and system based on machine learning

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102447697A (en) * 2010-11-15 2012-05-09 微软公司 Semi-private communication in open environments
CN103595688A (en) * 2013-11-04 2014-02-19 复旦大学 Visible light communication multiple access method and system based on carrierless amplitude/phase modulation
WO2015067140A2 (en) * 2013-11-07 2015-05-14 Huawei Technologies Co., Ltd. Method and apparatus for directly detected optical transmission systems based on carrierless amplitude-phase modulation

Family Cites Families (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9071363B2 (en) * 2013-09-12 2015-06-30 Futurewei Technologies, Inc. Optical transmitters with unbalanced optical sidebands separated by gaps
CN106100730B (en) * 2016-06-02 2018-11-13 复旦大学 Carrierless amplitude phase modulating system and modulator approach based on super Nyquist precoding
CN106453194A (en) * 2016-10-19 2017-02-22 华中科技大学 Method for transmitting signal by non-orthogonal multi-dimension non-carrier amplitude phase modulation technology
US20190052388A1 (en) * 2017-08-14 2019-02-14 Zte Corporation System and method for optical signal transmission
CN109217934B (en) * 2018-09-20 2020-04-21 哈尔滨工业大学（深圳） Polarization demultiplexing algorithm based on maximum likelihood independent component analysis method
CN111371493B (en) * 2020-03-08 2021-06-04 珠海复旦创新研究院 Machine learning multi-band carrier-free amplitude and phase modulation system based on independent element analysis

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102447697A (en) * 2010-11-15 2012-05-09 微软公司 Semi-private communication in open environments
CN103595688A (en) * 2013-11-04 2014-02-19 复旦大学 Visible light communication multiple access method and system based on carrierless amplitude/phase modulation
WO2015067140A2 (en) * 2013-11-07 2015-05-14 Huawei Technologies Co., Ltd. Method and apparatus for directly detected optical transmission systems based on carrierless amplitude-phase modulation

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
Non-linear Compensation of Multi-CAP VLC System Employing Clustering Algorithm-Based Perception Decision;Xingyu Lu等;《IEEE Photonics Journal》;20170904;第9卷(第5期);全文 *
时间混合调制无载波幅度相位可见光通信系统;石蒙等;《光通信研究》;20181010;全文 *

Also Published As

Publication number Publication date
CN111371493A (en) 2020-07-03
WO2021179845A1 (en) 2021-09-16

Similar Documents

Publication Publication Date Title
CN111371493B (en) Machine learning multi-band carrier-free amplitude and phase modulation system based on independent element analysis
Windpassinger Detection and precoding for multiple input multiple output channels
US20100104047A1 (en) Multiple-antenna space multiplexing system using enhancement signal detection
EP2200244A1 (en) Method and apparatus for multi-carrier frequency division multiplexing transmission
CN107395276A (en) A kind of visible light communication system of the ADO OFDM based on innovatory algorithm
CN104917599B (en) Transmission method when weighted score Fourier transformation expands in synchronization system
CN109889262A (en) A kind of orthogonal frequency division multiplexing free space optical communication method based on wavelet transformation
Shi et al. Experimental demonstration of OQAM-OFDM based MIMO-NOMA over visible light communications
CN111049586A (en) Pulse amplitude position modulation system based on amplitude reduction type probability forming
Wang et al. A novel algorithm for improving the spectrum efficiency of non-orthogonal multiband CAP UVLC systems
CN106100730B (en) Carrierless amplitude phase modulating system and modulator approach based on super Nyquist precoding
CN115296970A (en) Iterative orthogonal time-frequency-space waveform detection method based on element-by-element external information
CN107276726A (en) A kind of Massive MIMO FBMC beam space time coding downlink transmission methods
CN112398535B (en) Method for improving transmission capacity of non-orthogonal multiple access visible light communication based on probability shaping
US20100316107A1 (en) Frequency domain equalization method for continuous phase modulated signals
Chen et al. DC-balanced even-dimensional CAP modulation for visible light communication
Chi et al. Bandwidth-efficient visible light communication system based on faster-than-Nyquist pre-coded CAP modulation
Wang et al. Decision feedback kurtosis minimum crosstalk mitigation in super-Nyquist multiband CAP systems
CN113422646B (en) Zero value regression method and device for HACO-OFDM modulation system
Molteni et al. Joint OSC receiver for evolved GSM/EDGE systems
EP1931075B1 (en) Method of decoding of a received multidimensional signal
CN113507325A (en) IMDD optical communication system based on nonlinear differential coding and quadratic VNLE
CN102196111A (en) Short-wave two-path modulator-demodulator
WO2019127933A1 (en) Bidirectional qr decomposition and detection method and apparatus
CN116094880A (en) Visible light communication non-orthogonal multiple access system based on neural network precoding

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant