CLASSIFICATION DEVICE, CLASSIFICATION METHOD, AND PROGRAM

1 . A determination device that categorizes an installation environment of an optical cable, the determination device comprising:

a feature extraction unit that extracts a feature vector at each position of the optical cable from a vibration distribution in a longitudinal direction of the optical cable;

a calculation unit that calculates a discriminant function at each position of the optical cable from the feature vector and a weight vector corresponding to the installation environment, in need of categorization, of the optical cable; and

a determination unit that compares the discriminant function with teacher signals representing two states and designates a state of the nearer of the teacher signals as a state of the optical cable, at each position of the optical cable.

2 . The determination device according to claim 1 , wherein the feature extraction unit Fourier-transforms the vibration distribution from a waveform in a time domain to a spectrum waveform in a frequency domain, extracts frequencies of peaks and respective amplitudes of the peaks from the spectrum waveform and sorts the frequencies and the amplitudes into descending order of the amplitudes of the peaks to obtain the feature vector.

3 . The determination device according to claim 1 ,

wherein the feature extraction unit:

Fourier transforms the vibration distribution from a waveform in a time domain to a spectrum waveform in a frequency domain;

removes a fluctuation component from components having a period in a frequency axis direction of the spectrum waveform;

extracts frequencies of peaks and amplitudes of the peaks from the spectrum waveform from which the fluctuation component has been removed; and

configures the feature vector using the extracted frequencies of the peaks and amplitudes of the peaks.

4 . The determination device according to claim 3 ,

wherein the feature extraction unit removes the fluctuation component by:

logarithmizing absolute values of amplitudes of the spectrum waveform;

calculating a time waveform of components having a period in a frequency axis direction of the logarithmized spectrum waveform from the logarithmized spectrum waveform; and

extracting a time domain having a value smaller than a predetermined threshold period in the time waveform.

5 . The determination device according to claim 2 ,

wherein the weight vector is composed of coefficients for being multiplied by the frequencies and the respective amplitudes of the peaks correspondently, and

the calculation unit adds values obtained by multiplying the frequencies and the respective amplitudes of the peaks by the coefficients of the weight vector correspondently to obtain the discriminant function.

6 . The determination device according to claim 1 , further comprising:

a discriminant dictionary including a dictionary calculation unit and an update unit,

wherein the feature extraction unit extracts a known feature vector at each position of the optical cable from a vibration distribution in a longitudinal direction of the optical cable in each of the two states,

the dictionary calculation unit calculates a known discriminant function at each position of the optical cable from the known feature vector and the weight vector, and

the update unit calculates, at each position of the optical cable, an error between the known feature vector in each of the two states and the teacher signal representing a relevant state of the two states and updates the weight vector so as to reduce the error.

7 . The determination device according to claim 6 , wherein the discriminant dictionary settles the weight vector under condition that a square value of the error is a minimum and fluctuation of the error before and after update of the weight vector is equal to or less than a threshold.

8 . A determination method for categorizing an installation environment of an optical cable, the determination method comprising:

extracting a feature vector at each position of the optical cable from a vibration distribution in a longitudinal direction of the optical cable;

calculating a discriminant function at each position of the optical cable from the feature vector and a weight vector corresponding to the installation environment, in need of categorization, of the optical cable; and

comparing the discriminant function with teacher signals representing two states and designating a state of the nearer of the teacher signals as a state of the optical cable, at each position of the optical cable.

9 . The determination method according to claim 8 ,

wherein the vibration distribution is Fourier-transformed from a waveform in a time domain to a spectrum waveform in a frequency domain,

a fluctuation component is removed from components having a period in a frequency axis direction of the spectrum waveform,

frequencies of peaks and amplitudes of the peaks are extracted from the spectrum waveform from which the fluctuation component has been removed, and

the feature vector is configured using the extracted frequencies of the peaks and amplitudes of the peaks.

10 . The determination method according to claim 9 ,

wherein the fluctuation component is removed by:

logarithmizing absolute values of amplitudes of the spectrum waveform;

calculating a time waveform of components having a period in a frequency axis direction of the logarithmized spectrum waveform from the logarithmized spectrum waveform; and

extracting a time domain having a value smaller than a predetermined threshold period in the time waveform.

11 . The determination method according to claim 8 , further comprising:

extracting a known feature vector at each position of the optical cable from a vibration distribution in a longitudinal direction of the optical cable in each of the two states;

calculating a known discriminant function at each position of the optical cable from the known feature vector and the weight vector;

calculating, at each position of the optical cable, an error between the known feature vector in each of the two states and the teacher signal representing a relevant state of the two states; and

updating the weight vector so as to reduce the error.

12 . A program for instructing a computer to function as the determination device according to claim 1 .