 |
Claims  |
|
|
What is claimed is:
1. An adaptive equalizer to compensate for deterioration of an output signal from an object to be equalized, comprising:
characteristic detection means for detecting a characteristic of the object to be equalized to compensate for deterioration by using a known input signal and a first output signal of the object to be equalized corresponding to the known input
signal;
equalizing means for equalizing a second output signal of the object to be equalized corresponding to an unknown input signal using the characteristic detected by said characteristic detection means and for outputting an equalized output signal
corresponding to the second output signal from the object to be equalized; and
a characteristic detection control device, operatively connected to the object to be equalized, said equalizing means and said characteristic detection means, for initially supplying the known input signal and the first output signal to said
characteristic detection means while the known input signal is inputted to the object to be equalized and for subsequently supplying the equalized output signal from said equalizing means and the second output signal to said characteristic detection
means when the unknown input signal is inputted to the object to be equalized, to successively detect characteristics.
2. An adaptive equalizer to compensate for deterioration of an output signal from an object to be equalized, comprising:
characteristic detection means, formed of one of a neural network and a learning device, for detecting a characteristic of the object to be equalized to compensate for deterioration by using a known input signal and a first output signal of the
object to be equalized corresponding to the known input signal and for obtaining a characteristic detecting function by learning predetermined data presented thereto; and
equalizing means for equalizing a second output signal of the object to be equalized corresponding to an unknown input signal using the characteristic detected by said characteristic detection means.
3. The adaptive equalizer as set forth in claim 2, wherein said characteristic detection means detects one of a transfer function of the object to be equalized and values of a real part and an imaginary part of the transfer function.
4. The adaptive equalizer as set forth in claim 2, wherein said characteristic detection means detects values of an amplitude frequency characteristic and an envelope characteristic of a transfer function of the object to be equalized.
5. An adaptive equalizer to compensate for deterioration of an output signal from an object to be equalized, comprising:
characteristic detection means for detecting a characteristic of the object to be equalized to compensate for deterioration by using a known input signal and a first output signal of the object to be equalized corresponding to the known input
signal; and
equalizing means, formed of one of a neural network and a learning device, for equalizing a second output signal of the object to be equalized corresponding to an unknown input signal using the characteristic detected by said characteristic
detection means and for obtaining a characteristic detecting function by learning predetermined data presented thereto.
6. An adaptive equalizer to compensate for deterioration of an output signal from an object to be equalized, comprising:
characteristic detection means for detecting a characteristic of the object to be equalized to compensate for deterioration by using a known input signal and a first output signal of the object to be equalized corresponding to the known input
signal;
equalizing control means for outputting at least one control signal to control equalization of a second output signal of the object to be equalized corresponding to an unknown input signal using the characteristic detected by said characteristic
detection means;
equalizing means for equalizing the second output signal of the object to be equalized in accordance with the control signal output by said equalizing control means and for outputting an equalized output signal corresponding to the second output
signal from the object to be equalized; and
a characteristic detection control device, operatively connected to the object to be equalized, said equalizing means and said characteristic detection means, for initially supplying the known input signal and the first output signal to said
characteristic detection means while the known input signal is inputted to the object to be equalized and for subsequently supplying the equalized output signal from said equalizing means and the second output signal to said characteristic detection
means when the unknown input signal is inputted to the object to be equalized, to successively detect characteristics.
7. An adaptive equalizer to compensate for deterioration of an output signal from an object to be equalized, comprising:
characteristic detection means for detecting a characteristic of the object to be equalized to compensate for deterioration by using a known input signal and a first output signal of the object to be equalized corresponding to the known input
signal;
equalizing control means for outputting a plurality of control signals providing weighting signals to control equalization of a second output signal of the object to be equalized corresponding to an unknown input signal, in accordance with the
characteristic detected by said characteristic detection means; and
equalizing means, including a plurality of equalizing circuits operatively connected to receive the control signals, respectively, from said equalizing control means, for equalizing the second output signal of the object to be equalized in
accordance with the control signals output by said equalizing control means; and
means for producing an equalized output signal by adding the outputs of the equalizing circuits after weighting according to the control signals from said equalizing control means.
8. An adaptive equalizer to compensate for deterioration of an output signal from an object to be equalized, comprising:
characteristic detection means, formed of one of a neural network and a learning device, for detecting a characteristic of the object to be equalized to compensate for deterioration by using a known input signal and a first output signal of the
object to be equalized corresponding to the known input signal, and for obtaining a characteristic detecting function by learning predetermined data presented thereto;
equalizing control means for outputting at least one control signal to control equalization of a second output signal of the object to be equalized corresponding to an unknown input signal using the characteristic detected by said characteristic
detection means; and
equalizing means for equalizing the second output signal of the object to be equalized in accordance with the control signal output by said equalizing control means.
9. The adaptive equalizer as set forth in claim 8, wherein said characteristic detection means detects one of a transfer function of the object to be equalized and values of a real part and an imaginary part of the transfer function.
10. The adaptive equalizer as set forth in claim 8, wherein said characteristic detection means detects values of an amplitude frequency characteristic and an envelope characteristic of a transfer function of the object to be equalized.
11. An adaptive equalizer to compensate for deterioration of an output signal from an object to be equalized, comprising:
characteristic detection means for detecting a characteristic of the object to be equalized to compensate for deterioration by using a known input signal and a first output signal of the object to be equalized corresponding to the known input
signal;
equalizing control means, formed of one of a neural network and a learning device, for outputting at least one control signal to control equalization of a second output signal of the object to be equalized corresponding to an unknown input signal
using the characteristic detected by said characteristic detection means, and for obtaining a characteristic detecting function by learning predetermined data presented thereto; and
equalizing means for equalizing the second output signal of the object to be equalized in accordance with the control signal output by said equalizing control means.
12. An adaptive equalizer to compensate for deterioration of an output signal from an object to be equalized, comprising:
characteristic detection means for detecting a characteristic of the object to be equalized to compensate for deterioration by using a known input signal and a first output signal of the object to be equalized corresponding to the known input
signal;
equalizing control means for outputting at least one control signal to control equalization of a second output signal of the object to be equalized corresponding to an unknown input signal using the characteristic detected by said characteristic
detection means; and
equalizing means for equalizing the second output signal of the object to be equalized in accordance with the control signal output by said equalizing control means, said equalizing means a plurality of equalizing circuits, including one of a
neural network and learning devices for performing the equalizing after learning predetermined data presented thereto.
13. An adaptive equalizer to compensate for deterioration of an output signal from an object to be equalized, comprising:
characteristic detection means for detecting a characteristic of the object to be equalized to compensate for deterioration from a first output signal of the object to be equalized corresponding to a known input signal by using a detection
parameter corresponding to a signal pattern of the known input signal;
equalizing means for equalizing a second output signal of the object to be equalized corresponding to an unknown input signal using the characteristic detected by said characteristic detection means and for outputting an equalized output signal
corresponding to the second output signal from the object to be equalized; and
a characteristic detection control device, operatively connected to the object to be equalized, said equalizing means and said characteristic detection means, for initially supplying the known input signal and the first output signal to said
characteristic detection means while the known input signal is inputted to the object to be equalized and for subsequently supplying the equalized output signal from said equalizing means and the second output signal to said characteristic detection
means when the unknown input signal is inputted to the object to be equalized, to successively detect characteristics.
14. An adaptive equalizer to compensate for deterioration of an output signal from an object to be equalized, comprising:
characteristic detection means, formed of one of a neural network and a learning device, for detecting a characteristic of the object to be equalized to compensate for deterioration from a first output signal of the object to be equalized
corresponding to a known input signal by using a detection parameter corresponding to a signal pattern of the known input signal and for obtaining a characteristic detecting function by learning predetermined data presented thereto; and
equalizing means for equalizing a second output signal of the object to be equalized corresponding to an unknown input signal using the characteristic detected by said characteristic detection means.
15. The adaptive equalizer as set forth in claim 14, wherein said characteristic detection means detects one of a transfer function of said object to be equalized and values of a real part and an imaginary part of the transfer function.
16. The adaptive equalizer as set forth in claim 14, wherein said characteristic detection means detects values of an amplitude frequency characteristic and an envelope characteristic of a transfer function of the object to be equalized.
17. An adaptive equalizer to compensate for deterioration of an output signal from an object to be equalized, comprising:
characteristic detection means for detecting a characteristic of the object to be equalized to compensate for deterioration from a first output signal of the object to be equalized corresponding to a known input signal by using a detection
parameter corresponding to a signal pattern of the known input signal; and
equalizing means, formed of one of a neural network and a learning device, for equalizing a second output signal of the object to be equalized corresponding to an unknown input signal using the characteristic detected by said characteristic
detection means and for obtaining a characteristic detecting function by learning predetermined data presented thereto.
18. An adaptive equalizer to compensate for deterioration of an output signal from an object to be equalized, comprising:
characteristic detection means for detecting a characteristic of the object to be equalized to compensate for deterioration by using a known input signal and a first output signal of the object to be equalized corresponding to the known input
signal;
post-process means for storing characteristics detected previously by said characteristic detection means and for modifying the characteristic presently detected by said characteristic detection means using the characteristics detected previously
by said characteristic detection means to produce a modified result; and
equalizing means for equalizing a second output signal of the object to be equalized corresponding to an unknown input signal using the modified result of said post-process means.
19. The adaptive equalizer as set forth in claim 18,
wherein said equalizing means outputs an equalized output signal corresponding to the second output signal from the object to be equalized, and
wherein said adaptive equalizer further comprises a characteristic detection control device, operatively connected to the object to be equalized, said equalizing means and said characteristic detection means, for initially supplying the known
input signal and the first output signal to said characteristic detection means while the known input signal is inputted to the object to be equalized and for subsequently supplying the equalized output signal from said equalizing means and the second
output signal to said characteristic detection means when the unknown input signal is inputted to the object to be equalized, to successively detect characteristics.
20. The adaptive equalizer as set forth in claim 18, wherein said characteristic detection means detects the characteristic of the object to be equalized using a detection parameter in accordance with a signal pattern of the known input signal.
21. The adaptive equalizer as set forth in claim 18, wherein said characteristic detection means is one of a neural network and a learning device for obtaining a characteristic detecting function by learning predetermined data presented thereto.
22. The adaptive equalizer as set forth in claim 18, wherein said equalizing means is one of a neural network and a learning device for obtaining an equalizing function by learning predetermined data presented thereto.
23. The adaptive equalizer as set forth in claim 18, wherein said post-process means is a statistical process device.
24. The adaptive equalizer as set forth in claim 18, wherein said post-process means is one of a neural network and a learning device for obtaining a post-process function by learning predetermined data presented thereto.
25. The adaptive equalizer as set forth in claim 18, wherein said characteristic detection means detects one of a transfer function of the object to be equalized and values of a real part and an imaginary part of the transfer function.
26. The adaptive equalizer as set forth in claim 18, wherein said characteristic detection means detects values of an amplitude frequency characteristic and an envelope characteristic of a transfer function of the object to be equalized.
27. An adaptive equalizer to compensate for deterioration of an output signal from an object to be equalized, comprising:
characteristic detection means for detecting a characteristic of the object to be equalized to compensate for deterioration by using a known input signal and a first output signal of the object to be equalized corresponding to the known input
signal;
equalizing means for equalizing a second output signal of the object to be equalized corresponding to an unknown input signal using the characteristic detected by said characteristic detection means and for outputting first and second equalized
output signals respectively corresponding to the first and second output signals;
equalizing error detection and learning pattern storage means for detecting an equalizing error between the known input signal and the first equalized output signal and for storing the known input signal as a correct output signal, the first
equalized output signal and the characteristic detected by said characteristic detection means, to provide a learning pattern when the equalizing error is detected; and
learning control means for allowing said equalizing means to learn using the learning pattern when the object to be equalized is not outputting a signal.
28. The adaptive equalizer as set forth in claim 27, wherein said adaptive equalizer comprises an adjustment device, including
a characteristic detection device forming said characteristic detection means;
an equalizing device forming said equalizing means;
a recognition signal storage device, an equalizing error determination device, and a learning pattern storage device forming said equalizing error detection and learning pattern storage means; and
a learning control device forming said learning control means.
29. The adaptive equalizer as set forth in claim 27, wherein said adaptive equalizer further comprises a characteristic detection control device, operatively connected to the object to be equalized, said equalizing means and said characteristic
detection means, for initially supplying the known input signal and the first output signal to said characteristic detection means while the known input signal is inputted to the object to be equalized and for subsequently supplying the second equalized
output signal from said equalizing means and the second output signal to said characteristic detection means when the unknown input signal is inputted to the object to be equalized, to successively detect characteristics.
30. The adaptive equalizer as set forth in claim 27, wherein said characteristic detection means is one of a neural network and a learning device for obtaining a characteristic detecting function by learning predetermined data presented thereto.
31. The adaptive equalizer as set forth in claim 27, wherein said equalizing means is one of a neural network and a learning device for obtaining a characteristic detecting function by learning predetermined data presented thereto.
32. The adaptive equalizer as set forth in claim 27, wherein said characteristic detection means detects one of a transfer function of the object to be equalized and values of a real part and an imaginary part of the transfer function.
33. The adaptive equalizer as set forth in claim 27, wherein said characteristic detection means detects values of an amplitude frequency characteristic and an envelope characteristic of a transfer function of the object to be equalized. |
|
 |
|
 |
Claims |
|
|
|
Description |
|
|
The present invention relate to equalizers for compensating deterioration of signals over transmission lines of communication systems, and in particular, to adaptive equalizers for detecting
characteristics of transmission lines in radio and cable communications and for equalizing received signals by using the detected results.
As the information society grows, cable and radio communication networks are being widely constructed and serviced all over the world and information exchange using communication technologies is becoming increasingly dependent on this important
social infrastructure. In the information society, as well as voice signals or picture signals in conventional TV broadcasting, radio broadcasting, and telephone communication, information which remarkably affects our society such as business
information exchanged among companies, banking online information, and so forth are transmitted through communication networks. For such important information, accurate and error-free transmission means which are not affected by characteristics of
transmission lines must be established.
Particularly, in digital mobile communications for car telephones and so forth, a compensation technique for rapidly varying multi-path fading is becoming an important requirement. Thus, an adaptive equalizing technique for adaptively equalizing
deterioration of a signal over a transmission line in accordance with time-by-time variation of characteristics should be established.
A conventional radio transmission line comprises a radio device and a transmission line. The radio device further comprises a transmission filter, a reception filter, a modulator, a demodulator, a transmitter, and a receiver. The
characteristics of the transmission line vary depending on weather conditions, buildings, and so forth between the transmitter and the receiver. Causes of signal deterioration include linear and nonlinear distortion generated from constructional devices
of the system and two-wave interfered fading generated from the transmission line as linear distortion. Although the distortion generated from devices for mobile radio communication is stable with respect to time, the distortion of a propagation path
generated over the transmission line between the mobile station and the ground station becomes selective fading which varies with time. In accordance with this variation, the distortion of the signals over the transmission line should be adaptively
equalized.
FIG. 1A is a schematic showing the concept of a mobile communication. In the figure, a mobile communication is performed between a mobile station 1 and a ground station 2. The direct wave is directly propagated between the mobile station 1 and
the ground station 2. The reflected wave is reflected by an obstacle 3 such as a building, a mountain, ground, or the like. Thus, fading occurs between the direct wave and the reflected wave. In particular, when a mobile station moves at high speed in
mobile communication or the like, the amplitude ratio, propagation delay time, and phase difference between the direct wave and the reflected wave rapidly vary. Thus, the deterioration of the signals should be adaptively and rapidly equalized in
accordance with this variation.
FIG. 1B is a block diagram showing a construction of a transversal type equalizer of an equalizer as a related art. The transversal type equalizer demodulates a received signal supplied from the transmission line by using a demodulator 4.
Thereafter, the demodulated signal is continuously supplied to a tapped delay line 5. Each coefficient adjustment device 7 multiplies an output of each delay device 6 by a coefficient. An adder 8 adds output signals from the coefficient adjustment
devices 7 and outputs the result. By adjusting the coefficient of each coefficient adjustment device 7 in accordance with the degree of distortion of the propagation path and the constructional devices of the radio unit, the distortions are removed.
The conventional transversal type equalizer shown in FIG. 1B is effective against linear distortion. However, when nonlinear distortion occurs over a transmission line or when a nonlinear circuit is used in a demodulation system, received
signals are not properly equalized.
In addition, as a filter which is effective against nonlinear distortion, an equalizer using a neural network is known. However, since such an equalizer requires a long learning time, it cannot perform an adaptive process. Therefore, it is not
effective against variation of the characteristics of a transmission line. Thus, when the characteristics of the transmission line rapidly vary, as in multi-path fading, this equalizer is not effective.
SUMMARY OF THE INVENTION
Therefore, an object of the present invention is to provide a function for detecting characteristics of a transmission line and equalizing a signal by using the detected result and thereby adaptively equalizing both linear and nonlinear
distortions of the signal.
A feature of the present invention resides in an adaptive equalizer, comprising characteristic detection means for detecting a characteristic of an object to be equalized by using an output signal in accordance with a known input signal of the
object to be equalized as a deterioration compensation of the output signal; and equalizing means for equalizing an output signal of the object to be equalized in accordance with an unknown input signal by using the detected result of the characteristic
detection means.
BRIEF DESCRIPTION OF THE DRAWINGS
FIG. 1A is a schematic drawing of two-wave interfered fading in a mobile communication system in accordance with the present inventions;
FIG. 1B is a block diagram showing an example of a transversal type equalizer.
FIG. 2 is a block diagram for explaining the theory of the first and third embodiments of the invention;
FIG. 3 is a block diagram for explaining the theory of the second embodiment of the invention;
FIG. 4 is a block diagram for explaining the theory of the fourth embodiment of the invention;
FIG. 5 is a block diagram for explaining the theory of the fifth embodiment of the invention;
FIG. 6 is a block diagram showing the construction of an adaptive equalizer in accordance with the first embodiment of the invention;
FIG. 7A is a block diagram showing the construction of an adaptive equalizer in accordance with the first embodiment of the invention;
FIG. 7B is a schematic diagram showing a model of the propagation path in accordance with the first embodiment of the invention;
FIG. 8 is a block diagram showing the construction of a modulation system of an embodiment of the adaptive equalizer of FIG. 7A;
FIG. 9 is a block diagram showing the construction of a demodulation system of an embodiment of the adaptive equalizer of FIG. 7A;
FIG. 10 is a schematic diagram showing the packet format of an embodiment in accordance with the present invention;
FIG. 11 is a schematic diagram showing the characteristic detection neural network of an embodiment of the adaptive equalizer of FIG. 6;
FIG. 12 is a block diagram showing the construction of an embodiment of an input pattern generation portion in the characteristic detection neural network;
FIG. 13 is a block diagram showing the construction of a training sequence detection portion;
FIG. 14 is a block diagram showing the construction of a training sequence memory;
FIG. 15 is a diagram showing input signals of the characteristic detection neural network of FIG. 11;
FIG. 16 is a table showing transmission line parameters in a simulation for generating learning data;
FIG. 17 is a schematic diagram used in describing the necessity to learn all combinations of 3 bits on each channel of I and Q;
FIG. 18 is a table showing an example of learning data of the characteristic detection neural network;
FIG. 19 are tables used to describe countermeasures against waveform distortion for learning data of a characteristic detection neural network;
FIG. 20 is a schematic diagram showing an equalizing neural network of an embodiment of the adaptive equalizer of FIG. 6;
FIG. 21 is a block diagram for describing a data input control system of the equalizing neural network;
FIG. 22 is a timing chart showing an operation of an embodiment of the equalizing neural network;
FIG. 23 is a block diagram for describing an operation delay compensation system of the characteristic detection neural network;
FIG. 24A is a table and 24B is a schematic diagram together used to describe an example of learning data in the equalizing neural network;
FIG. 25 is a flow chart of an embodiment of the output signal equalizing process in accordance with the first invention;
FIG. 26 is a schematic diagram showing sampling points in a signal equalizing experiment in the first embodiment of the invention;
FIG. 27 is a bar chart showing the result of the signal equalizing experiment in the first embodiment of the invention;
FIG. 28 is a block diagram showing the construction of an embodiment of a feedback type adaptive equalizer in accordance with the first embodiment of the invention;
FIG. 29 is a block diagram showing the construction of the characteristic detection control apparatus;
FIG. 30 is a block diagram showing the construction of a received signal FIFO;
FIGS. 31A and 31B are timing charts showing a signal equalizing process in accordance with the embodiment of FIG. 28;
FIG. 32 is a block diagram for explaining the theory of an adaptive equalizer in accordance with the second invention;
FIG. 33 is a block diagram showing the construction of an embodiment of an adaptive equalizer in accordance with the second invention;
FIG. 34 is a schematic diagram showing a control neural network of an embodiment of the adaptive equalizer of FIG. 33;
FIG. 35 is a table showing an example of learning data for use in a control neural network;
FIG. 36 is a table showing an example of learning data of a control neural network where amplitude ratio and phase difference are varied;
FIG. 37 is a schematic diagram showing an equalizing neural network of an embodiment of an adaptive equalizer of FIG. 33;
FIG. 38 is a block diagram showing the construction of an embodiment of a feedback type equalizer in accordance with the second embodiment of the invention;
FIG. 39 is a block diagram showing the construction of an adaptive equalizer in accordance with the third embodiment of the invention;
FIG. 40 is a block diagram showing the construction of an embodiment of an adaptive equalizer in accordance with the third embodiment of the invention;
FIG. 41 is a schematic diagram showing a characteristic detection neural network of an embodiment of the adaptive equalizer of FIG. 40;
FIG. 42 is a block diagram showing the construction of an embodiment of the characteristic detection neural network parameter management portion;
FIG. 43 is a block diagram showing the construction of the weight control portion;
FIG. 44 is a schematic diagram showing an example of learning data of the characteristic detection neural network;
FIG. 45 is a block diagram showing the construction of a feedback type adaptive equalizer in accordance with the third embodiment of the invention;
FIG. 46 is a block diagram showing the construction of an adaptive equalizer in accordance with the fourth embodiment of the invention;
FIG. 47 is a block diagram showing the construction of an embodiment of an adaptive equalizer in accordance with the fourth embodiment of the invention;
FIG. 48 is a block diagram showing the construction of an embodiment of an adaptive equalizer using characteristic detection parameters according to a bit pattern in accordance with the fourth embodiment of the invention;
FIG. 49 is a schematic diagram showing an embodiment of a post-process neural network;
FIG. 50 is a table showing an example of learning data of the post-process neural network;
FIG. 51 is a flow chart showing an embodiment of a signal equalizing process in accordance with the fourth embodiment of the invention;
FIG. 52 is a block diagram showing the construction of an embodiment of a feedback type adaptive equalizer (No. 1);
FIG. 53 is a block diagram showing the construction of an embodiment of a feedback type adaptive equalizer (No. 2);
FIG. 54 is a block diagram showing the construction of a characteristic storage device and a statistical process device in accordance with the fourth embodiment of the invention;
FIG. 55 is a block diagram showing the construction of an adaptive equalizer in accordance with the fifth embodiment of the invention;
FIG. 56 is a block diagram showing the construction of an embodiment of the adaptive equalizer in accordance with the fifth embodiment of the invention;
FIG. 57 is a schematic diagram showing an example of a determining of an equalizing error;
FIG. 58 is a schematic showing an example of a learning pattern generated by the determination of an equalizing error;
FIG. 59 is a block diagram showing the construction of an embodiment of the learning control device of FIG. 56;
FIG. 60 is a block diagram showing the construction of an embodiment of a feedback type adaptive equalizer in accordance with the fifth embodiment of the invention; and
FIG. 61 is a block diagram showing the construction of an embodiment where an adjustment device of the adaptive equalizer is separately provided.
DETAILED DESCRIPTION OF THE INVENTION
FIGS. 2 to 5 are block diagrams for explaining theories of the present invention. These figures are block diagrams of adaptive equalizers for detecting characteristics of an object to be equalized where a known input signal is inputted and for
compensating the output signal. The object to be equalized is for example a transmission line.
FIG. 2 is a block diagram for explaining a theory of the first and third embodiment of the invention. In the first embodiment of the invention, a characteristic detection unit 11 detects characteristics such as a transfer function of an object
to be equalized 10, for example a transmission line, which is to be equalized to compensate deterioration of an output signal. The characteristics are detected by comparing a known input signal with an output signal of the object to be equalized 10
corresponding to the known input signal in the characteristic detection unit 11. An equalizing unit 12 equalizes an output signal of the object to be equalized 10 corresponding to an unknown input signal by using the detected result of the
characteristic detection unit 11.
FIG. 3 is a block diagram for explaining a theory of the second embodiment of the invention. In FIG. 3, a characteristic detection unit 11 detects characteristics of an object to be equalized 10 in the same manner as in the first embodiment of
the invention. An equalizing control unit 14 uses the detected result of the characteristic detection unit 11 and outputs a control signal for equalizing an output signal of the object to be equalized 10 corresponding to an unknown input signal. An
equalizing unit 15 equalizes the output signal of the object to be equalized 10 corresponding to an unknown input signal in accordance with the control signal output by the equalizing control unit 14. The equalizing unit 15 comprises for example a
plurality of equalizing circuits and a weighting addition circuit for weighting and adding the output signals of the equalizing circuit. The equalizing control unit 14 outputs a control signal 13 corresponding to each output signal of the plurality of
equalizing circuits. Thus, the output signals of the plurality of equalizing circuits are weighted and the sum thereof is outputted as an equalized output signal.
FIG. 2 also explains the theory of the third embodiment of the invention. As in the first embodiment of the invention, in the third embodiment of the invention, the characteristic detection unit 11 is for example a neural network. The
characteristic detection unit 11 detects characteristics such as a transfer function of the object to be equalized 10, for example, a transmission line, which is to be equalized to compensate for deterioration of an output signal, by using an output
signal in accordance with a known input signal of the object to be equalized 10. However, in the first embodiment of the invention, values of the internal state of the neural network as the characteristic detection unit 11, for example the weight of
internal linkage, are constant regardless of the signal pattern of the known input signal. In contrast, in the third embodiment of the invention, the weight of the internal linkage is switched in accordance with the signal pattern of the known input
signal. Thereby, the neural network forming characteristic detection means 11 effectively executes a learning process. As in the first embodiment of the invention, an equalizing unit 12 equalizes an output signal of the object to be equalized 10
corresponding to an unknown input signal by using the detected result of the characteristic detection means 11 and outputs an equalized output signal.
FIG. 4 is a block diagram for explaining the theory of the fourth embodiment of the invention. In FIG. 4, a characteristic detection unit 11 is for example a neural network, as in the third embodiment of the invention. The characteristic
detection unit 11 detects characteristics of an object to be equalized 10. However, unlike in the third embodiment of the invention, the weight as an internal state value of the neural network is not switched in accordance with the signal pattern of an
input signal. The characteristics are detected while the weight of the internal linkage is not changed.
A post-process unit 16 successively stores the result detected by the characteristic detection unit 11 by using output signals of an object to be equalized 10 which are-outputted in a time series manner and performs a post-process on the detected
result of the characteristics being stored, for example, an average value of transfer functions. The post-process unit 16 is constructed of for example a shift register and a neural network. An equalizing unit 17 equalizes an output signal of the
object to be equalized 10 by using an Output signal of the post-process unit 16 as the detected result of the characteristics of the object to be equalized 10. The equalizing unit 17 is also constructed of for example a neural network.
FIG. 5 is a block diagram for explaining the theory of the fifth embodiment of the invention. The operations of a characteristic detection unit 11 and an equalizing unit 12 of the fifth embodiment of the invention are the same as those of the
first embodiment of the invention.
An equalizing error detection and learning pattern storage unit 18 compares a known input signal to an object to be equalized 10 with an output signal of an equalizing unit 12. When this detection and storage unit 18 detects an equalizing error,
it stores the output signal, the detected result of the characteristic detection unit 11 and the known input signal which is a correct output signal to be output from the equalizing unit 12, as a learning pattern.
A learning control means 19 causes the equalizing unit 12, for example a neural network, to learn by using a learning pattern stored in the equalizing error detection and learning pattern storage unit 18 while the object to be equalized 10 is not
outputting a signal, for example, while the transmission line as the object to be equalized 10 is not being used.
In the present invention, when data is transmitted over a transmission line in a packet format, a known bit train named a training sequence or a recognition signal is placed at the beginning of a packet. This bit train is used as a known input
signal for detecting characteristics of the object to be equalized 10, for example the transmission line. By using the relation between this input signal and the output signal of the object to be equalized 10, the characteristic detection unit 11, for
example the neural network, detects the characteristics of the transmission line, for example the real part and the imaginary part of a transfer function thereof.
In the first embodiment of the invention shown in FIG. 2, by using for example the real part and the imaginary part of a transfer function detected by a neural network forming characteristic detection means 11, the equalizing unit 12 equalizes an
output signal of the transmission line corresponding to an unknown input signal. The equalizing unit 12 is also formed of for example a neural network.
In the second embodiment of the invention shown in FIG. 3, by using an output signal of a neural network forming characteristic detection means 11, a control signal for equalizing an output signal of an object to be equalized 10 corresponding to
an unknown input signal is outputted. This control signal provides a weight signal to weight each output signal of a plurality of equalizing circuits forming equalizing means 15. The weighted results of output signals of the plurality of equalizing
circuits are added and outputted as an equalized output signal. Each equalizing circuit is also constructed of a neural network.
In the third embodiment of the invention shown in FIG. 2, a neural network forming characteristic detection means 11 switches weight of the neural network as a characteristic detection parameter, in accordance with a bit pattern of a training
sequence which is a known signal and detects the real part and the imaginary part of a transfer function. Thereby, an output signal of the transmission line corresponding to an unknown input signal is equalized.
In the fourth embodiment of the invention shown in FIG. 4, a bit train of a training sequence is shifted for example bit by bit and input to a characteristic detection unit 11. The detected result of each bit train is stored in, for example, a
shift register. For the stored result, a post-process, for example an averaging process, is performed. Thereafter, by using the result of the post-process, an output signal of the object to be equalized 10 corresponding to an unknown input signal is
equalized. Thereby, in comparison with the construction where the detected result of characteristics is used one time, the accuracy of the detection of characteristics can be improved.
In the fifth embodiment of the invention shown in FIG. 5, as in the first embodiment of the invention, a characteristic detection unit 11 detects characteristics of an object to be equalized 10. By using the detected result, an output signal of
the object to be equalized 10 corresponding to an unknown input signal is equalized. At the same time, an equalizing error detection and learning pattern storage unit 18 compares a known input signal of an object to be equalized 10 with an output signal
of an equalizing unit 12. When the storage means 18 detects an equalizing error, it stores the output signal, the detected result of the characteristic detection unit 11, and a correct output signal of the object to be equalized 10 as a learning
pattern. For example, while no communication takes place on the transmission line, the equalizing unit 12 performs a learning process by using this learning pattern.
In the above description, a training sequence placed at the beginning of a packet is used as the known input signal of the object to be equalized 10. However, when fading takes place at very high speed, even while one packet is being received,
the characteristics of the transmission line may vary. To meet such a situation, according to the present invention, even after a signal in accordance with a training sequence is received, characteristics can be continuously detected. To continuously
detect characteristics assuming that an equalized output signal of the adaptive equalizer corresponding to an unknown input signal is correct (as a known signal), a characteristic detection control device is provided in a preceding stage of the
characteristic detection unit 11 shown in FIGS. 2 to 5 (the first to fifth embodiments of the invention).
The characteristic detection control device supplies a known input signal and an output signal of an object to be equalized 10 to a characteristic detection unit 11 while a training sequence corresponding to the known input signal is inputted to
the object to be equalized 10. After the input of the training sequence is completed, while an unknown input signal 8, i.e., transmission data is being inputted to the object to be equalized 10, the characteristic detection control device supplies both
an output signal of the equalizing means 12 as a signal instead of the known input signal and an output signal of the object to be equalized 10 to the characteristic detection unit 11. Thereby, the characteristic detection unit 11 continuously detects
the characteristics.
As described above, according to the present invention, by detecting transfer characteristics of an object to be equalized, for example, a transfer function and by using the detected result, an output signal can be equalized in accordance with
the transfer characteristics which vary each time.
EMBODIMENTS
FIG. 6 is a block diagram showing a construction of an adaptive equalizer in accordance with the first embodiment of the invention. In FIG. 6, the adaptive equalizer 20 comprises a characteristic detection device 22 for detecting a transfer
characteristic, for example, transfer function H (.omega.) of an object to be equalized 21; and an equalizing device 23 for equalizing an output signal of the object to be equalized 21 by using the transfer function characteristic being detected.
In FIG. 6, the characteristic detection device 22 and the equalizing device 23 are for example neural networks. The characteristic detection device 22 learns with the transfer function of the object to be equalized 21 using a teacher signal.
The equalizing device 23 learns with an input signal of the object to be equalized 21 as a teacher signal.
FIG. 7A is a block diagram showing the construction of the adaptive equalizer in accordance with the first embodiment of the invention. In FIG. 7A, the object to be equalized 21 of FIG. 6 comprises a modulation system 26, a propagation path 27,
and a demodulation system 28. The adaptive equalizer 25 comprises an input pattern generation portion 24, which will be described later in detail; a characteristic detection neural network 29, which is equivalent to the characteristic detection device
22 of FIG. 6; an equalizing neural network 30, which is equivalent to the equalizing device 23 of FIG. 6; and a threshold process portion 31. The characteristic detection neural network 29 detects a real part X | | |
