Claims
- 1. A learning type signal recording and reproducing apparatus comprising:
- a reference signal generating means for generating predetermined reference waveform signals and applying them to a recording media for recordation thereby,
- a signal detection means for detecting recorded signals from the recording media, the recorded signals corresponding to the reference waveform signals recorded by the media, and
- a conversion rule generating means receiving detected recorded signals from the signal detection means and the corresponding reference waveform signals from the reference signal generating means, and generating conversion rules based on a relationship between the detected recorded signals and the reference waveform signals to correctly reproduce signals recorded on the recording media.
- 2. A learning type signal recording and reproducing apparatus according to claim 1, wherein
- the conversion rule generating means includes:
- a serial signal sampler for sampling as serial signals, signal changes at each point on a transition axis of the detected recorded signals, and
- a learning type similarity calculation unit for automatically generating a conversion rule by learning a relationship between the signal changes at each point on the transition axis sampled by the serial signal sampler and the reference waveform signals generated by the reference signal generating means.
- 3. A learning type signal recording and reproducing apparatus according to claim 2, wherein
- the learning type similarity calculation unit includes a plurality of recognition units, the plurality of recognition units each including
- a signal input section for the serial signals, and
- a path selector which has at least one path input terminal and at least one path output terminal and which selects a path in accordance with the entered serial signals; and
- whereby the plurality of recognition units are formed in a network to have a multi-layer hierarchy.
- 4. A learning type signal recording and reproducing apparatus according to claim 3, wherein
- a recognition unit of a preceding layer of an output layer of the network receives reference waveform signals in place of the serial signals to the signal input section and includes:
- a path learning device which varies connection strengths between the at least one path input terminal and the at least one path output terminal according to the conversion rule, and
- an output results judging device for detecting a maximum output recognition unit of the output layer of the network.
- 5. A learning type signal recording and reproducing apparatus according to claim 2, wherein
- the serial signal sampler includes:
- a present signal sampler for sampling output signals at a specified point on the transition axis of the detected recorded signals, and
- a change rate sampling section for calculating a difference between the signals sampled by the present signal sampler and output signals before said specified point.
- 6. A learning type signal recording and reproducing apparatus according to claim 5, wherein
- the change rate sampling section includes:
- a memory for storing present signals at each point on the transition axis of the detected recorded signals sampled by the signal sampler, one by one in a specified range, and
- a differential calculation unit for calculating a difference between the signal sampled by the present signal sampler and the signal stored in the memory.
- 7. A waveform signal equalizing apparatus comprising:
- serial signal sampling means for sampling signals at each point on a transition axis of a waveform signal which is outputted from a specified signal system,
- waveform signal recognition means for obtaining an output result representing a change to a signal caused by the specified signal system and sampled by the serial signal sampling means using pattern recognition, and
- equalizing signal generating means for generating equalizing signals based on the output result of the waveform signal recognition means, wherein
- the waveform signal recognition means contains waveform signal values as a plurality of recognition categories corresponding to a predetermined reference teaching signal,
- the plurality of recognition categories represent the predetermined reference teaching signal in a condition unmodified by the specified signal system, and
- the waveform signal recognition means has a learning type similarity calculation means which learns a learnable relationship between serial signals sampled by the serial signal sampling means and the plurality of recognition categories, whereby
- when the predetermined reference teaching signal is inputted to the specified signal system, the learning type calculation means learns the learnable relationship, and
- when unknown signals are inputted to the specified signal system, the waveform signal recognition means, based on the learnable relationship learned by the similarity calculation means, outputs the output result for the serial signals sampled at each point on the transition axis.
- 8. A waveform signal equalizing apparatus according to claim 7, wherein
- the learning type similarity calculation means has
- a signal input section for inputting serial signals, and
- a path selector which has at least one path input terminal and at least one path output terminal and selects a path in accordance with inputted serial signals;
- recognition units are formed in a network to have a multi-layer hierarchy;
- a recognition unit of a preceding layer of an output layer of the network receives the reference teaching signals; and
- a path learning device which varies connection strengths between the at least one path input terminal and the at least one path output terminal according to the learned relationship.
- 9. A waveform signal equalizing apparatus according to claim 7, wherein the equalizing signal generating means includes
- a similarity inner divider for calculating an inner division similarity ratio to the recognition categories that are provided by the waveform signal recognition means, and
- a signal value synthesizer for calculating a product of the sum of the inner division ratio and signal values corresponding to each recognition category.
- 10. A waveform signal equalizing apparatus according to claim 7, wherein
- the equalizing signal generating means detects a maximum similarity value for recognition categories provided by the waveform signal recognition means.
- 11. A waveform signal equalizing apparatus according to claim 7, further comprising
- a reference teaching signal generator for generating reference teaching signals including a multiplicity of signal values as recognition categories, and
- a header sampler for sampling a header of input waveform signals, wherein
- the learning type similarity calculation unit learns a relationship between the serial sampled signals at each point on the transaction axis of signals in the header sampled by the header sampler and the reference teaching signals generated by the reference teaching signal generator.
- 12. A waveform signal equalizing apparatus according to claim 11, wherein
- the header sampler comprises a waveform input detector for detecting a head of a waveform signal and a gate pulse generator for generating gate pulses for a specified time after the waveform input detector detects the head of a waveform signal.
- 13. A waveform signal equalizing apparatus according to claim 11, wherein
- the header sampler includes a header specifying signal detector for detecting a header specifying signal from outside, and
- a specified range sampler for sampling a specified range of a waveform signal in accordance with an output of the header specifying signal detector.
- 14. A waveform signal equalizing apparatus according to claim 13, wherein
- the specified range sampler includes a header specifying pulse generator for generating gate pulses during a specified range of the header signal in accordance with an output of the header specifying signal detector.
- 15. A waveform generation method comprising the steps of:
- applying a teaching signal to a system;
- obtaining a waveform signal from said system;
- sampling said waveform signal at each point on a transition axis to obtain serial signals;
- applying said teaching signal and said serial signals to a pattern recognition means in order to obtain a relationship between said teaching signal and said serial signals; and
- determining original signals applied to said system based upon waveform signals obtained from said system and said relationship between said teaching signal and said serial signals.
- 16. A waveform generation method according to claim 15, wherein
- said serial signals include a last applied sampled signal and differences between sampled signals including current and past sampled signals.
Priority Claims (2)
Number |
Date |
Country |
Kind |
5-86153 |
Apr 1993 |
JPX |
|
5-257120 |
Oct 1993 |
JPX |
|
Parent Case Info
This application is a divisional of application Ser. No. 08/221,905 filed Apr. 1, 1994, now abandoned.
US Referenced Citations (14)
Foreign Referenced Citations (1)
Number |
Date |
Country |
3-118605 |
May 1991 |
JPX |
Divisions (1)
|
Number |
Date |
Country |
Parent |
221905 |
Apr 1994 |
|