TOP > 外国特許検索 > Identification device, identification method, and identification processing program

Identification device, identification method, and identification processing program

外国特許コード F120006092
整理番号 S2008-0225
掲載日 2012年1月6日
出願国 アメリカ合衆国
出願番号 81100609
公報番号 20100287133
公報番号 8321368
出願日 平成21年1月16日(2009.1.16)
公報発行日 平成22年11月11日(2010.11.11)
公報発行日 平成24年11月27日(2012.11.27)
国際出願番号 JP2009050529
国際公開番号 WO2009093525
国際出願日 平成21年1月16日(2009.1.16)
国際公開日 平成21年7月30日(2009.7.30)
優先権データ
  • 特願2008-012678 (2008.1.23) JP
  • 2009JP050529 (2009.1.16) WO
発明の名称 (英語) Identification device, identification method, and identification processing program
発明の概要(英語) There are provided an identification device, an identification method and an identification processing program, which are capable of significantly reducing a processing burden.
An identification device 1 can judge the magnitude relation between an occurrence probability value of a class 0 and an occurrence probability value of a class 1 from the magnitude relation between gkupper and gklower.
Hence, it can be identified which one of the classes 0 and 1 is applicable to observed data D1 with a simple arithmetic processing.
Accordingly, a complicated and heavy-burden arithmetic processing of an exponential function can be avoided for obtaining the occurrence probability values of the classes 0 and 1, enabling the processing burden to be significantly reduced.
従来技術、競合技術の概要(英語) BACKGROUND ART
In recent years, as a technique for a sensor to sense an identification target to identify what the identification target is like based on observed data obtained from the sensor, there is known such a technique that a certain probability distribution model is assumed to identify the identification target according to the Bayes' decision rule (refer to, e.g., patent document 1).
Under the assumption that the observed data each follow a single Gaussian distribution, the exponential functions multiplied by a certain constant K: K exp(-z), are compared to one another to thereby enable pattern recognition.
This pattern recognition can be realized by the comparison between the numbers of (ln K-z) produced by applying a logarithm to the function and hence there is no need to calculate an exponential function in an identification device.
It is to be noted herein that ln K is a constant.
Here, a single Gaussian distribution is unsuitable to data that follow a multi-modal distribution and therefore has disadvantages of limited applications.
This problem with the multi-modal distribution, however, can be improved by introducing Gaussian mixture distributions expressed by the following formula which means a weighted sum of Gaussian distributions.
(Equation image 1 not included in text)
The patent document 1: Japanese unexamined patent application publication No. 2005-267570

特許請求の範囲(英語) [claim1]
1. An identification device for classifying observed data based on parameters of Gaussian mixture distributions on the assumption that a distribution of said observed data follows Gaussian mixture distributions, said identification device comprising: a power of two multiplier for calculating the following formulae:
hk,nupper=Kk,n2-[zk,n log2 e] [Formula 40]
(where [zk, n log2 e] denotes an integer part of zk, n log2 e)
hk,nlower=hk,nupper2-1 [Formula 41]
using a group zk, n of variables obtained based on respective feature vectors of a plurality of said observed data (k denotes 1 or 0 indicating a class, and n denotes a distribution number of a Gaussian distribution assumed in each class) and a group Kk, n of constants obtained based on said parameter of Gaussian mixture distributions;
an accumulator for calculating the following formulae 42, 43
(Equation image 14 not included in text)
(Nk denotes the number of Gaussian mixture distributions of a class k) using said hk, nupper and said hk, nlower; and a comparator for comparing said gkupper and said gklower to classify said observed data, using g1upper <= g0lower and g0upper <= g1lower.
[claim2]
2. The identification device according to claim 1, wherein said identification device is equipped with a storage unit in which the following formulae are stored,
B[i]=2<2-i [Formula 44]

B[i]-1=2<2-i [Formula 45]
(where i=0, 1 . . . , L, wherein L is a positive integer that is arbitrarily set)
and when said gkupper and said gklower are compared to each other to prove impossible to determine the magnitude relation between said gkupper and said gklower, said value of i is changed into a value of (i+1) and then it is judged whether a value at the ith decimal place in a fraction part of zk, n log2 e is 1 or 0, and when the value at the ith decimal place is proved to be 1 as a result, said hk, nupper is multiplied by said B[1] in said formula 44 and thus said hk, nupper is updated to thereby calculate said gkupper, whereas when the value at the ith decimal place is proved to be 0, said hk, nlower is multiplied by said B[1]-1 in said formula 45 and thus said hk, nlower is updated to thereby calculate said gklower.
[claim3]
3. The identification device according to claim 2, wherein said identification device is equipped with an averaging processor which calculates the following formula:
gkpseudo=2-1{gklowerB[L]-1+gkupper} [Formula 46]
when said value of i is said L to thereby classify said observed data by using g1pseudo<g0pseudo and g0pseudo<g1pseudo.
[claim4]
4. An identification method for classifying observed data based on parameters of Gaussian mixture distributions on the assumption that a distribution of said observed data follows Gaussian mixture distributions, said identification method comprising steps of: performing power of two multiplication for calculating the following formulae:
hk,nupper=Kk,n2-[zk,n log2 e] [Formula 47]
(where [zk, n log2 e] denotes an integer part of zk, n log2 e)
hk,nlower=hk,nupper2-1 [Formula 47]
using a group zk, n of variables obtained based on respective feature vectors of a plurality of said observed data (k denotes 1 or 0 indicating a class, and n denotes a distribution number of a Gaussian distribution assumed in each class) and a group Kk, n of constants obtained based on said parameter of Gaussian mixture distributions; performing accumulation for calculating the following formulae:
(Equation image 15 not included in text)
(where Nk denotes the number of Gaussian mixture distributions of a class k) using said hk, nupper and said hk, nlower, which have been calculated in said step of performing power of two multiplication; and comparing said gkupper and said gklower, which have been calculated in said step of performing accumulation to classify said observed data, using g1upper <= g0lower and g0upper <= g1lower.
[claim5]
5. The identification method according to claim 4, wherein said identification method comprises a step of a refining process in which when said gkupper and said gklower are compared to each other to prove impossible to determine the magnitude relation between said gkupper and said gklower, a value of i (where i=0, 1 . . . , L, and L is a positive integer that is arbitrarily set) is changed into a value of (i+1) and then it is judged whether the ith decimal place in a fraction part of zk, n log2 e is 1 or 0 and as a result, when the ith decimal place of zk, n log2 e is proved to be 1, said hk, nupper is multiplied by the following formula:
B[i]=2-2-i [Formula 51]
and thus said hk, nupper is updated to thereby calculate said gkupper, whereas when the ith decimal place of zk, n log2 e is proved to be 0, said hk, nlower is multiplied by the following formula
B[i]-1=22-i [Formula 52]
and thus said hk, nlower is updated to thus calculate said gklower.
[claim6]
6. The identification method according to claim 5, wherein said identification method is equipped with a step of an averaging process in which when a value of said i is said L, the following formula:
gkpseudo=2-1{gklowerB[L]-1+gkupper} [Formula 53]
is calculated to classify said observed data by using g1pseudo <= g0pseudo and g0psuedo <= g1pseudo.
[claim7]
7. An identification processing program stored on a non-transitory computer storage medium for classifying observed data based on parameters of Gaussian mixture distributions on the assumption that a distribution of said observed data follows Gaussian mixture distributions, said identification processing program allowing a computer to execute the steps of: performing power of two multiplication for calculating the following formulae:
hk,nupper=Kk,n 2[-zk,n log2 e] [Formula 54]
(where [zk, n log2 e] denotes an integer part of Zk, n log2 e)
hk,nlower=hk,nupper 2-1 [Formula 55]
by using a group zk, n of variables obtained based on respective feature vectors of a plurality of said observed data (k denotes 1 or 0 indicating a class, and n denotes a distribution number of a Gaussian distribution assumed in each class) and a group Kk,n of constants obtained based on said parameter of Gaussian mixture distributions; performing accumulation for calculating the following formulae:
(Equation image 16 not included in text)
(Nk denotes the number of Gaussian mixture distributions of a class k) by using said hk,nupper and said hk,nlower, which have been calculated in said step of performing power of two multiplication; and comparing said gkupper and said gklower to classify said observed data by using g1upper <= g0lower g0upper g1lower.
[claim8]
8. The identification processing program according to claim 7, wherein said identification processing program comprises a step of a refining process in which when said gkupper and said gklower are compared to each other to prove impossible to determine the magnitude relation between said gkupper and said said gklower, said value of i (i=0, 1 . . . , L, and L is a positive integer arbitrarily set) is changed into a value of (i+1) and then it is judged whether a value at the ith decimal place in a fraction part of zk, n log2 e is 1 or 0, and as a result, when the value at the ith decimal place of zk, n log2 e is proved to be 1, said hk, nupper is multiplied by the following formula:
B[i]=2-2-i [Formula 58]
and thus said hk, nupper is updated to thus calculate said gkupper, whereas when said value at the ith decimal place of zk, n log2 e is proved to be 0, said hk, nlower is multiplied by the following formula:
B[i]-1=22-i [Formula 59]
and thus said hk, nlower is updated to thus calculate said gklower.
[claim9]
9. The identification processing program according to claim 8, wherein said identification processing program comprises a step of an averaging process in which when a value of said i is said L, the following formula:
gkpseudo=2-1{gklowerB[L]-1+gkupper} [Formula 60]
is calculated to classify said observed data by using g1pseudo <= g0pseudo and g0psuedo <= g1pseudo.
  • 発明者/出願人(英語)
  • MURAMATSU SHOGO
  • WATANABE HIDENORI
  • NIIGATA UNIVERSITY
国際特許分類(IPC)
米国特許分類/主・副
  • 706/52
  • 382/224
  • 706/12
ライセンスをご希望の方、特許の内容に興味を持たれた方は、下記「問合せ先」までお問い合わせください。

PAGE TOP

close
close
close
close
close
close