DCT-based adaptive metric learning model using asymptotic local information measure

Takami Satonaka; Takaaki Baba; Takayuki Chikamura; Tatsuo Otsuki; Teresa H. Meng

DCT-based adaptive metric learning model using asymptotic local information measure

Takami Satonaka^*, Takaaki Baba, Takayuki Chikamura, Tatsuo Otsuki, Teresa H. Meng

^*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

1 Citation (Scopus)

Abstract

We present an adaptive metric learning vector quantization procedure based on the discrete-cosine transform (DCT) for accurate face recognition used in multimedia application. Since the set of learning samples may be small, we employ a mixture model of prior distributions. The model selection method, which minimizes the cross entropy between the real distribution and the modeled one, is presented to optimize the mixture number and local metric parameters. The structural risk minimization is used to facilitate an asymptotic approximation of the cross entropy for models of fixed complexity. We also provide a formula to estimate the model complexity derived from the minimum description length criterion. The structural risk minimization method proposed achieves an recognition error rate of 2.29% using the ORL database, which is better than previously reported numbers using the Karhunen-Loeve transform convolution network, the hidden Marcov model and the eigenface model.

Original language	English
Title of host publication	Neural Networks for Signal Processing - Proceedings of the IEEE Workshop
Place of Publication	Piscataway, NJ, United States
Publisher	IEEE
Pages	521-530
Number of pages	10
Publication status	Published - 1997
Externally published	Yes
Event	Proceedings of the 1997 7th IEEE Workshop on Neural Networks for Signal Processing, NNSP'97 - Amelia Island, FL, USA Duration: 1997 Sept 24 → 1997 Sept 26

Other

Other	Proceedings of the 1997 7th IEEE Workshop on Neural Networks for Signal Processing, NNSP'97
City	Amelia Island, FL, USA
Period	97/9/24 → 97/9/26

ASJC Scopus subject areas

Signal Processing
Software
Electrical and Electronic Engineering

Cite this

Satonaka, T, Baba, T, Chikamura, T, Otsuki, T & Meng, TH 1997, DCT-based adaptive metric learning model using asymptotic local information measure. in Neural Networks for Signal Processing - Proceedings of the IEEE Workshop. IEEE, Piscataway, NJ, United States, pp. 521-530, Proceedings of the 1997 7th IEEE Workshop on Neural Networks for Signal Processing, NNSP'97, Amelia Island, FL, USA, 97/9/24.

@inproceedings{0ab76e1b0c9a4c8ebeb641a7289537ed,

title = "DCT-based adaptive metric learning model using asymptotic local information measure",

abstract = "We present an adaptive metric learning vector quantization procedure based on the discrete-cosine transform (DCT) for accurate face recognition used in multimedia application. Since the set of learning samples may be small, we employ a mixture model of prior distributions. The model selection method, which minimizes the cross entropy between the real distribution and the modeled one, is presented to optimize the mixture number and local metric parameters. The structural risk minimization is used to facilitate an asymptotic approximation of the cross entropy for models of fixed complexity. We also provide a formula to estimate the model complexity derived from the minimum description length criterion. The structural risk minimization method proposed achieves an recognition error rate of 2.29% using the ORL database, which is better than previously reported numbers using the Karhunen-Loeve transform convolution network, the hidden Marcov model and the eigenface model.",

author = "Takami Satonaka and Takaaki Baba and Takayuki Chikamura and Tatsuo Otsuki and Meng, {Teresa H.}",

year = "1997",

language = "English",

pages = "521--530",

booktitle = "Neural Networks for Signal Processing - Proceedings of the IEEE Workshop",

publisher = "IEEE",

note = "Proceedings of the 1997 7th IEEE Workshop on Neural Networks for Signal Processing, NNSP'97 ; Conference date: 24-09-1997 Through 26-09-1997",

}

TY - GEN

T1 - DCT-based adaptive metric learning model using asymptotic local information measure

AU - Satonaka, Takami

AU - Baba, Takaaki

AU - Chikamura, Takayuki

AU - Otsuki, Tatsuo

AU - Meng, Teresa H.

PY - 1997

Y1 - 1997

N2 - We present an adaptive metric learning vector quantization procedure based on the discrete-cosine transform (DCT) for accurate face recognition used in multimedia application. Since the set of learning samples may be small, we employ a mixture model of prior distributions. The model selection method, which minimizes the cross entropy between the real distribution and the modeled one, is presented to optimize the mixture number and local metric parameters. The structural risk minimization is used to facilitate an asymptotic approximation of the cross entropy for models of fixed complexity. We also provide a formula to estimate the model complexity derived from the minimum description length criterion. The structural risk minimization method proposed achieves an recognition error rate of 2.29% using the ORL database, which is better than previously reported numbers using the Karhunen-Loeve transform convolution network, the hidden Marcov model and the eigenface model.

AB - We present an adaptive metric learning vector quantization procedure based on the discrete-cosine transform (DCT) for accurate face recognition used in multimedia application. Since the set of learning samples may be small, we employ a mixture model of prior distributions. The model selection method, which minimizes the cross entropy between the real distribution and the modeled one, is presented to optimize the mixture number and local metric parameters. The structural risk minimization is used to facilitate an asymptotic approximation of the cross entropy for models of fixed complexity. We also provide a formula to estimate the model complexity derived from the minimum description length criterion. The structural risk minimization method proposed achieves an recognition error rate of 2.29% using the ORL database, which is better than previously reported numbers using the Karhunen-Loeve transform convolution network, the hidden Marcov model and the eigenface model.

UR - http://www.scopus.com/inward/record.url?scp=0031351688&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0031351688&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:0031351688

SP - 521

EP - 530

BT - Neural Networks for Signal Processing - Proceedings of the IEEE Workshop

PB - IEEE

CY - Piscataway, NJ, United States

T2 - Proceedings of the 1997 7th IEEE Workshop on Neural Networks for Signal Processing, NNSP'97

Y2 - 24 September 1997 through 26 September 1997

ER -

DCT-based adaptive metric learning model using asymptotic local information measure

Abstract

Other

ASJC Scopus subject areas

Other files and links

Fingerprint

Cite this