JPH05282727A - Method for reproducing magnetooptic signal and magnetooptic disk device - Google Patents

Abstract

PURPOSE:To provide the magnetooptic disk device which is suitable for high- density recording by improving the S/N of a reproduced signal from a magnetooptic disk and improving deterioration in the signal quality due to interference between codes. CONSTITUTION:The magnetooptic disk device is equipped with a laser light source 4, an objective 7 which converges a light beam and irradiates a disk medium 3 having information tracks, a beam splitter 6 which splits the reflected light from said disk 3, a polarizing prism 14, and plural photodetectors 17 and 19 which photoelectrically convert the reflected light and transmitted light from the polarizing prism. Two information signals S1 and S2 obtained from the outputs of the photodetectors 17 and 19 are passed through different neural networks 21 and 22 and the addition output or subtraction output of two obtained outputs is used as the reproduced information signal by the magnetooptic signal reproducing method and magnetooptic disk device.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a magneto-optical disk device for recording / reproducing information using an optical disk, and more particularly to a method of detecting an information signal from a magneto-optical disk and an information signal detecting means.

[0002]

2. Description of the Related Art An example of a conventional magneto-optical disk device will be described with reference to FIG. FIG. 7 is a diagram showing an example of a conventional magneto-optical disk device. In FIG. 1, recording of information on the optical disk medium 3 having a number of information tracks arranged in a circular pattern is performed by the magneto-optical recording method using the optical head 1. Further, the reading of information signals from the optical disc medium is also performed by using the optical head 1. The optical head 1 includes a light source 4 such as a semiconductor laser, a collimator lens 5 that transforms the light beam from the light source 4 into a parallel light beam, a light beam emitted from the light source 4 such as the semiconductor laser, and the optical disc medium 3.
A polarization beam splitter 6 and the like for separating the reflected light flux from the.

In FIG. 7, P is a light spot condensed by the objective lens 7. In the optical head 1, the objective lens 6 is moved in the radial direction of the magneto-optical disk 3 to form the light spot P. To move in a direction perpendicular to the information track (hereinafter referred to as a tracking direction), and the objective lens 6 is moved in a direction perpendicular to the surface of the magneto-optical disk 3. A focusing actuator (not shown) for focusing the spot P on the surface of the optical disk is also provided.

The reflected light from the disk medium 3 is reflected by the beam splitter 6 and enters the beam splitter 9. The light beam reflected by the beam splitter 9 enters the condenser lens 10, is condensed, and enters the photodetector 11. A tracking error signal and a focusing error signal are detected from the output of this photodetector. The light flux passing through the beam splitter 9 is incident on the half-wave plate 13,
The polarization plane is changed and the light enters the analyzer 13 such as a polarization prism. The light flux is split into two by the polarization prism 13, and approximately the same amount of light flux enters the condenser lens 18 and the condenser lens 15. The light flux passing through the condenser lens 15 enters a photodetector 16 and an output terminal 17 of the photodetector 16 is provided.
An information signal (1) (hereinafter, referred to as S1) is output to. The light flux passing through the condenser lens 18 enters a photodetector 19, and an information signal (2) (hereinafter referred to as S2) is output to the output terminal 20 of the photodetector 19.

The two outputs of the optical head 1, that is, the information signal (1) S1 and the information signal (2) S2 are both outputs according to the information recorded on the optical disk medium 3. These two signals are input to the differential amplifier 28,
The differential output is output to the output terminal 29 by the information signal (3) (hereinafter,
It is written as S3. ). This differential output S3
Is used as a reproduction information signal from the light disc 3.

As described above, the information tracks are formed on the reflecting film surface of the magneto-optical disk 3 in a concentric or spiral shape, and signals are detected on the information tracks by utilizing the light diffraction effect. The large number of pits may be pre-engraved in an uneven shape, and information may be written or read by the magneto-optical recording method. The information recorded in the uneven pits includes
For example, there are management data such as disks and information tracks, and servo signals. Other information is usually recorded and reproduced by a magneto-optical recording method.

The information signal recorded / reproduced by the magneto-optical recording method has a small output level of the reproduction signal and a small signal-to-noise ratio (hereinafter referred to as S / N) (bad). A difference detection method is generally used for the detection of (3). That is, the DC components of the signal S1 and the signal S2 are substantially equal, but the polarities of the information components (AC components) are opposite to each other. Therefore, if the magneto-optical disk device operates perfectly and the signals S1 and S2 are ideally detected, the information component (AC component) of the signal S3 is doubled, which is the main feature of the magneto-optical disk device. The light amount fluctuation noise, which is noise, is canceled because it has the same polarity component, and the S / N is improved accordingly.

However, in an actual magneto-optical disk device, the effect of the differential detection is lowered due to the imperfections of the recording / reproducing system, and the S / N ratio of the signal S3 is deteriorated. There is a problem that there will be a loss of. In the conventional magneto-optical disk device in which the differential detection method works well,
The deterioration of the S / N is a major factor of the deterioration of the S / N remaining in the disk device. The imperfection of the system is a deviation from an ideal state, and the cause thereof is mainly in the optical system. For example, the phase difference of the ½ wavelength plate 13 is out of order, and the polarization prism 1
4 is the deviation of the polarized light separation characteristic from the ideal state, the disorder of the polarized light due to the lens, and the asymmetrical characteristics of the two photodetectors 16 and 19.

The recording / reproducing system of an optical disk can be easily understood from the OTF (Optical Transfer Performance) of the objective lens, the response characteristics of the recording material, etc.
It can be considered as a kind of low pass filter. When trying to increase the recording density, the recording / reproducing system is used up to near the upper limit of the frequency band, and when the adjacent marks are read, the reproduced waveforms of the respective marks are likely to interfere with each other, resulting in a read error. This phenomenon is called intersymbol interference. Intersymbol interference caused by band limitation based on the spatial frequency characteristic of this optical system is a major factor limiting the recording capacity of the magneto-optical disk.

[0010]

In the magneto-optical disk device, the S / N ratio of the reproduced information signal caused by the imperfections of the recording / reproducing system is improved, and the deterioration of the signal quality due to the intersymbol interference is improved. It is to provide a magneto-optical disk device suitable for high density recording.

[0011]

A laser light source, an objective lens for focusing a light beam from the light source on a disk medium having an information track and irradiating it as a light spot, and a reflected light from the disk medium are separated. In a magneto-optical disk device comprising a beam splitter, an analyzer, and a plurality of photodetectors for photoelectrically converting the reflected light and the transmitted light from the analyzer, the output from the plurality of photodetectors is obtained. A magneto-optical disk reproducing method and a magneto-optical disk device in which the two information signals obtained are passed through different neural nets, and the addition output or the subtraction output of the obtained two outputs is used as a reproduction information signal. Further, the present invention is a magneto-optical disk device in which the output of a neural network using two information signals obtained from the outputs of the plurality of photodetectors as input signals is used as a reproduction information signal, and further, the magneto-optical recording method is applied to the disk medium. In a magneto-optical signal reproducing method and a magneto-optical disk device, signal processing is performed on a magneto-optical signal recorded by the method and information from a pit previously recorded by the shape of the unevenness by using different neural network systems. is there.

[0012]

The laser light flux reflected by the magneto-optical disk is split into approximately equal light fluxes by the polarizing prism. The divided luminous fluxes are input to different photodetectors. The outputs S1 and S2 of the two photodetectors are signals in which signal components (AC components) whose polarities are opposite to each other are superimposed on DC components and noise of substantially equal levels. The two signals S1 and S2 are input to individual neural nets (hereinafter, referred to as NN). This NN is a NN that has been hardwareized based on the synapse coupling coefficient determined by performing simulation and learning by a computer outside the magneto-optical disk device. To work. The outputs of the two NNs are subtracted by a subtractor, and the result is finally used as a reproduction information signal from the disc. In the NN simulation, a signal close to the signal waveform at the time of recording is used as a teacher signal, and an ideal output is obtained including various characteristics of the recording / reproducing system of the magneto-optical disk device. Since it is performed in step S1, the output signal with good S / N and suppressed intersymbol interference is obtained.

[0013]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS In the magneto-optical disk device of the present invention, the reflected light from the magneto-optical disk is divided into two by a polarizing prism, and an information signal (1) (hereinafter referred to as S1) obtained from these luminous fluxes, The information signal (2) (hereinafter referred to as S2) is input to each neural network, and the N
The two information signals waveform-equalized by N are applied to a subtractor, and the output signal of the subtractor is the information signal (3).
(Hereinafter referred to as S3).

A magneto-optical signal reproducing method and a magneto-optical disk device according to a first embodiment of the present invention will be described below with reference to FIGS. 1 is a diagram showing a first embodiment of a magneto-optical disk device of the present invention, FIG. 2 is a conceptual diagram of a waveform equivalent circuit using a neural network, and FIG. 3 is an explanatory diagram of each unit of the neural network shown in FIG. is there. In FIG. 1, FIG.
The same components and the same functions and effects and the same signals are denoted by the same reference numerals, and the description thereof will be omitted.

The optical head 1 and the magneto-optical disk 3 shown in FIG.
Are the same as the optical head 1 and the magneto-optical disk 3 shown in FIG. 7, respectively. The laser light generated from the light source 4 is
It is collimated into a parallel light flux through the collimator lens 5, is converged through the beam splitter 6 and the objective lens 7, and is emitted to the reflecting surface of the disc 3. The reflected light from the disc 3 is
The light is incident on the beam splitter 6 through the objective lens 7, is changed in direction and is separated from the incident light, and is incident on the condenser lenses 15 and 18 via the beam splitter 9, the half-wave plate 13, and the polarization prism 14, respectively. To be done. Light fluxes from the condenser lenses 15 and 18 are incident on photodetectors 16 and 19, respectively, and output terminals 1 of the photodetectors 16 and 19 are input.
Information signals (1) S1 and information signals (2) S2 are included in 7 and 20.
Are output respectively.

The output signals S1 and S2 of the photodetector are input to the signal synthesizing circuit 2, and the output of this signal synthesizing circuit 2 has a drawback due to the imperfections of the recording / reproducing system of the magneto-optical disk device. The compensated information signal (3) S3 is output. The signal synthesis circuit 2 includes two NNs, 21, 22.
And a subtractor 25, and the output signal S of the photodetector
1 and S2 are input to the neural nets 21 and 22, respectively, and high precision information signals 23 and 24 are output to the output terminals of these NNs, respectively. These signals 2
3 and 24 are applied to the subtractor 25, and the difference between the signals 23 and 24 is output to the output terminal 26 of the subtractor 25 as the information signal (3). This information signal (3) is a highly accurate signal in which a defect due to the imperfections of the recording / reproducing system of the magneto-optical disk device is compensated, and is used as a highly accurate reproduced information signal.

The structures of the neural nets 21 and 22 shown in FIG. 1 are as shown in FIGS. The neural network as shown in FIGS. 2 and 3 is described in detail in Japanese Patent Application No. 31952 of 1991 filed by the present applicant. The outline of the NN will be described below with reference to FIGS. 2 and 3. FIG. 2 shows a conceptual diagram of a waveform equivalent circuit including four delay elements 41 to 44 in addition to the NN21 configured by hardware. The waveform equalizing circuit is a circuit for processing a rectangular wave-shaped input signal si having a waveform distortion by the NN to obtain an output so that is as close to an ideal rectangular wave as possible. The waveform equivalent circuit can be realized without using the NN, but it has been experimentally confirmed that a very good waveform equivalent circuit can be realized by using the NN.

In FIG. 2, delay elements 41 to 44 are delay elements each having a predetermined delay time. NN21
Is composed of an input layer composed of buffer amplifiers 51 to 55, an intermediate layer composed of neurons 61 to 63, and an output layer composed of neuron 71, and is an input signal to each delay element 41 to 44. Are buffer amplifiers 51 to 5
5, and the outputs of these buffer amplifiers are input to the neurons 61 to 63 in the intermediate layer through the three individual resistors each having a resistance value determined by the synapse weight (synapse coupling coefficient). In the figure, a small circle indicates the existence of this synaptic coupling coefficient, and a symbol W indicates the existence of a synaptic weight.

FIG. 3 is a diagram showing unit connection of the neural network shown in FIG. In FIG. 3, a unit (U) is a conversion system which has a nonlinear input / output characteristic in which a brain neuron is modeled as an engine, and the neural network 21 is a combination of many units connected by variable weights W. It is the body. FIG. 3 is a diagram showing unit connection of the neural network shown in FIG. The unit U is a conversion system configured by a nonlinear function F (X) that outputs an output value Y = F (X) for an input value X, for example, a nonlinear input / output characteristic. As the input value, the sum of the results obtained by multiplying the output value of the previous layer (or the input value to the neural network) by the independent variable weight W is input, and this input value is nonlinearly converted. It is output to the rear layer.

More specifically, FIG. 3 is a diagram showing the i-th unit combination of the n-th layer forming the layered neural network, and the input value X to the i-th unit of the n-th layer is shown.
(N, i) is the strength of the coupling between the j-th unit of the (n-1) -th layer and the i-th unit of the n-th layer in the output of each unit of the (n-1) -th layer which is the previous layer. Weight W (n, i, j)
It is configured to be the sum of those multiplied by. Then, the input value X (n, i) is input to the non-linear function F (X), non-linearly converted to Y (n, i) = F {X (n, i)}, and output to the unit in the next layer. Is configured.

The variable weight W (n, i, j) represents the coupling strength of the units (U). The weight (n, i, j) is changed, converged, and learned. The desired neural net is constructed. Here, the synapse weight is previously determined by learning of the back propagation method using an ideal signal as a teacher signal. Since there are positive and negative synapse weights, each neuron is configured with hardware so as to correspond to positive and negative synapse weights. Since this NN has a small number of neurons, it can be easily realized as hardware.

In the present invention, the learning for determining the hardware of the NN is performed by the delay element and the NN constructed by software in the computer outside the magneto-optical disk device, and the learning for each delay element is performed. Input signal is NN
Are also simultaneously input to each unit of the input layer, and a teacher signal is created by calculation based on a predetermined calculation algorithm. At the same time, the output of the NN is compared with the teacher signal, and each synapse coupling coefficient in the NN is corrected based on the deviation amount. The process is performed on a number of data to determine synaptic coupling coefficients. On the basis of the determined synapse coupling coefficient, a hardware NN is provided in the magneto-optical disk device.

By the way, in the magneto-optical disk apparatus of the present invention, the NN itself incorporated in the magneto-optical disk apparatus does not have a learning function in order to make the NN compatible with high-speed processing. Next, the neural nets 21 and 22 shown in FIG.
An example of how to determine the synaptic coupling coefficient of is described. First, simulation and learning are performed by an external computer. For this purpose, the signals S1 and S2 shown in FIG. 1 are fetched by a computer, learning is carried out by the back propagation method to determine the synapse coupling coefficient, and the actual hardware circuit constant is determined based on this result. ..

As an actual procedure, first, the NN21 is designed by the procedure as described in the above-mentioned prior application by the present applicant. The output waveform of the signal input to and processed by the NN 21 is waveform-equalized. Here, the main points of the design of the NN21 will be described. In general, a back-propagation NN design requires a teacher signal and a student signal. When NN is applied to a signal transmission system such as an optical disk, the teacher signal is an original waveform and the student signal is a signal deteriorated in the transmission system.

Here, by taking advantage of the characteristics of the magneto-optical disk device capable of recording and reproducing, learning is performed by using the original signal as a teacher signal and the reproduced signal as a student signal. The two signals are digitized, taken into a computer, and learned by the back propagation method to determine the configuration of the NN. The length of the captured signal needs to be some length for the purpose of improving various patterns of the signal. Good results can be obtained by capturing signals from approximately 10 to several tens of marks.

Further, the following are two main points in the learning. 1) As the teacher signal, a waveform obtained by blunting an original waveform (rectangular wave) through a low-pass filter is used. This means that it is necessary for the NN to have an extremely wide-band frequency characteristic in order to make the reproduced waveform, which is frequency-limited by the optical system, into a rectangular wave all at once by using the NN. It is an excessive load. That is, learning when a rectangular wave is used as a teacher signal does not proceed easily, and when learning is a non-linear process, so-called local minimum often falls and learning often stops. However, by using the waveform in which the teacher signal is blunted, good learning can be performed without imposing an unnecessarily excessive demand on the NN or a constraint condition.

The band of the low-pass filter for the original waveform may be sufficiently wider than the band of the optical system. As a guide, good results can be obtained if there is a band which is 3 to 5 times or more the signal band. When passing through such a low-pass filter, the intersymbol interference of the original waveform only slightly increases, so that the teacher signal is sufficiently good. 2) In learning, it is necessary to correctly match the phases of the original signal and the reproduced signal. There are various methods for phase matching, but here, taking advantage of the characteristics of the magneto-optical disk device capable of recording and reproducing, a specific pattern for cueing the original signal is prepared and the same pattern is searched for with the reproduced signal. I decided to match the phase.

The output signal waveform of the NN21, which has been hardware-implemented based on the result learned as described above, is a waveform that is strongly waveform-equivalent. Next, the NN22 is designed. As the teacher signal in this case, the NN21
A modified version of the output of is used. That is, a signal obtained by inverting only the component of the magneto-optical signal at the output of the NN21 is used as the teacher signal. As a result, the completed output of the NN 22 is subjected to waveform equalization and becomes a signal in which the imperfections of the detection system are compensated. In the learning of the NN22, it can be said that two points, that is, the effect of equalizing the waveform and the effect of compensating for imperfections between the information signal (1) and the information signal (2), are learned as the constraint conditions. .. In addition, in FIG. 1, an adder may be used instead of the subtractor 25. In this case, NN21, 2
It goes without saying that the synaptic coupling coefficient of 2 only changes.

Next, a magneto-optical signal reproducing method and a magneto-optical disk apparatus according to a second embodiment of the present invention will be described with reference to FIGS. The difference between the first embodiment and the second embodiment of the present invention is that
The signal synthesis circuit 2 shown in FIG. 1 is the signal synthesis circuit 2 shown in FIG.
Since it is only A, the description of parts other than this signal combining circuit will be omitted. FIG. 4 is a diagram showing a second embodiment of the signal synthesizing circuit, and FIG. 6 is a diagram showing the configuration of the neural network shown in FIG.

In FIG. 4, the information signal (1) S1 and the information signal (2) S2 are input to the NN 40 which constitutes the signal synthesizing circuit 2A, and from this NN, the information signal (3) is input.
Is used as a highly accurate reproduction information signal. The NN shown in FIG. 6 is configured such that there are two NNs shown in FIG. 2. When the signal S1 is input, the delay elements 41, 42,
43 and 44 are sequentially delayed. The input signal to this delay element is simultaneously supplied to the buffer amplifier 5 which constitutes the NN input layer.
1 to 55, and the signal S2 is similarly delayed by the delay elements 45, 46, 47 and 48 when input. The input signal to this delay element is also simultaneously input to the buffer amplifiers 56 to 60 forming the NN input layer.

The buffer amplifier of the input layer of the NN 40 and the neurons 61, 62 and 63 forming the intermediate layer are connected by a predetermined synaptic coupling coefficient (the synaptic coupling coefficient is not shown). The neurons 61 to 63 in the intermediate layer and the neuron 71 in the output layer are similarly connected with synaptic coupling coefficients. In actual hardware, these synaptic coupling coefficients are determined by simulation and learning by an external computer, and based on the result, the neurons of each layer are coupled by a resistance having a resistance value.

Next, a third embodiment of the magneto-optical signal reproducing method and the magneto-optical disk device according to the present invention will be described with reference to FIG. The difference between the third embodiment and the first embodiment of the present invention is that the signal synthesizing circuit 2 shown in FIG. 1 is replaced by the signal synthesizing circuit 2B shown in FIG. The description is omitted. FIG. 5 is a diagram showing a third embodiment of the signal combining circuit.

As described above, the magneto-optical disk may be provided with prepits formed as uneven pits. The purpose is to show address information and the like.
Alternatively, the disk may include a ROM section (reproduction-only section) and a RAM section (write-only section), and the ROM section may be formed by prepits. The principle of reproduction of a prepit formed by utilizing the diffraction phenomenon of light is different from the principle of reproduction of a magneto-optical signal which is mainly used by detecting a change in the polarization state of light. Therefore, the action of the reproduction signal of the pre-pit is different from that of the magneto-optical signal including the nonlinearity of the reproduction process.

In the conventional magneto-optical disk device, the magneto-optical signal is detected by the differential detection method at the time of reproduction as described above. On the other hand, the pre-pit signal appears in the form of a change in the amount of light. Therefore, the reproduction of the pre-pit is obtained by adding the two outputs, which is the same as when the signal is detected without using the analyzer. Therefore, the reproduction of the pre-pit may use another detector. In the embodiment shown in FIG. 5, the reproduction principle is different between the prepit and the magneto-optical signal. Therefore, by providing two NN signal processing systems corresponding to each, optimum reproduction signals for both signals are provided. Can be obtained.

In FIG. 5, an NN 70 is an NN for processing a magneto-optical signal, which is similar to the NN 40 shown in FIG. 4, and an NN 50 is an NN for processing a signal from the prepit. The output of the adder 27 to which the signals S1 and S2 are applied is applied. The switching circuit SW is a circuit for selectively outputting only one of the output of the NN50 and the output of the NN70 according to a control signal.

[0036]

According to the method of reproducing a magneto-optical signal and the magneto-optical disk apparatus of the present invention, the noise component which could not be removed by the conventional technique is canceled and the S / of the reproduced information signal in the magneto-optical disk apparatus is canceled. It is possible to provide a magneto-optical disk device suitable for high-density recording by improving N and improving deterioration of signal quality due to intersymbol interference.

[Brief description of drawings]

FIG. 1 is a diagram showing a first embodiment of a magneto-optical disk device of the present invention.

FIG. 2 is a conceptual diagram of a waveform equivalent circuit using a neural network.

FIG. 3 is a diagram showing unit connection of the neural network shown in FIG.

FIG. 4 is a diagram showing a second embodiment of the signal combining circuit.

FIG. 5 is a diagram showing a second embodiment of the signal combining circuit.

6 is a diagram showing a configuration of the neural network shown in FIG.

FIG. 7 is a diagram showing an example of a conventional magneto-optical disk device.

[Explanation of symbols]

1 ... Optical head, 3 ... Magneto-optical disk, 4 ...
Laser light source 2, 2A, 2B ... Signal combining circuit 5 ... Collimating lens 6, 9 ... Beam splitter 7 ... Objective lens
13 ... 1/2 wave plate 10, 15, 18 ... Condensing lens 11, 16, 1
9 ... Photodetector 14 ... Polarizing prism (analyzer) 21, 22, NN, 40, 50 ... Neural net 25 ... Subtractor (adder or subtractor) 27 ... Adder P ... Optical spot

Claims (5)

[Claims] 1. A laser light source, an objective lens for converging a light beam from the light source onto a disk medium having an information track and irradiating it as a light spot, and a beam splitter for separating reflected light from the disk medium. In a magneto-optical disk device comprising an analyzer and a plurality of photodetectors for photoelectrically converting reflected light and transmitted light from the analyzer, two pieces of information obtained from outputs of the plurality of photodetectors. 2 signals obtained by passing the signals through different neural nets
A reproduction method of a magneto-optical signal, wherein an addition output or a subtraction output of the two outputs is used as a reproduction information signal. 2. A laser light source, an objective lens for condensing a light beam from the light source onto a disk medium having an information track and irradiating it as a light spot, and a beam splitter for separating reflected light from the disk medium. In a magneto-optical disk device comprising an analyzer and a plurality of photodetectors for photoelectrically converting reflected light and transmitted light from the analyzer, two pieces of information obtained from outputs of the plurality of photodetectors. 2 signals obtained by passing the signals through different neural nets
A magneto-optical disk device characterized in that an addition output or a subtraction output of two outputs is used as a reproduction information signal. 3. A laser light source, an objective lens for focusing a light beam from the light source on a disk medium having an information track and irradiating it as a light spot, and a beam splitter for separating reflected light from the disk medium. In a magneto-optical disk device comprising an analyzer and a plurality of photodetectors for photoelectrically converting reflected light and transmitted light from the analyzer, two pieces of information obtained from outputs of the plurality of photodetectors. A magneto-optical disk device characterized in that a reproduction information signal is output from the neural network through a neural network using the signal as an input signal. 4. A laser light source, an objective lens for condensing a light beam from the light source onto a disk medium having an information track and irradiating it as a light spot, and a beam splitter for separating reflected light from the disk medium. In a magneto-optical disk device equipped with an analyzer and a plurality of photodetectors for photoelectrically converting reflected light and transmitted light from the analyzer, a magneto-optical signal recorded by a magneto-optical recording method and unevenness Information from the pits recorded in advance according to the shape,
A method for reproducing a magneto-optical signal, characterized by performing signal processing using separate neural network systems. 5. A laser light source, an objective lens for condensing a light beam from the light source onto a disk medium having an information track and irradiating it as a light spot, and a beam splitter for separating reflected light from the disk medium. In an optical disc device including an analyzer and a plurality of photodetectors for photoelectrically converting reflected light and transmitted light from the analyzer, an information signal recorded by a magneto-optical recording method and an uneven shape are used in advance. A magneto-optical disk device characterized in that information signals from recorded pits are subjected to signal processing using separate neural network systems.