The design method for a vector codebook using a variable weight and employing an improved splitting method

개선된 미세분할 방법과 가변적인 가중치를 사용한 벡터 부호책 설계 방법

  • Cho, Che-Hwang (Dept. of Electrical and Electronic Eng., Donshin Univ.)
  • 조제황 (東新大學敎 電氣電子工學科)
  • Published : 2002.07.01

Abstract

While the conventional K-means algorithms use a fixed weight to design a vector codebook for all learning iterations, the proposed method employs a variable weight for learning iterations. The weight value of two or more beyond a convergent region is applied to obtain new codevectors at the initial learning iteration. The number of learning iteration applying a variable weight must be decreased for higher weight value at the initial learning iteration to design a better codebook. To enhance the splitting method that is used to generate an initial codebook, we propose a new method, which reduces the error between a representative vector and the member of training vectors. The method is that the representative vector with maximum squared error is rejected, but the vector with minimum error is splitting, and then we can obtain the better initial codevectors.

벡터 부호책 설계에 사용되는 기존 K-means 알고리즘은 모든 학습반복에서 고정된 가중치를 적용하는데 반해 제안된 방법은 학습반복마다 가변되는 가중치를 적용한다. 초기 학습반복에서는 새로운 부호벡터를 얻기 위해 수렴영역을 벗어나는 2 이상의 가중치를 사용하고, 이 값이 클수록 가변 가중치를 적용하는 학습반복을 줄임으로써 우수한 부호책을 설계할 수 있다. 초기 부호책 설계에 사용되는 미세분할 방법을 개선하기 위하여 소속 학습벡터와 대표벡터간의 오차를 줄이는 방법을 사용한다. 즉 자승오차가 최대인 대표벡터를 제외시키고 최소인 대표벡터를 미세분할함으로써 초기 부호벡터로 대체될 보다 적절한 대표벡터를 얻을 수 있다.

Keywords

References

  1. Y. Linde, A.Buzo, and R.M.Gray, 'An algorithm for vector quantizer design, ' IEEE Trans. Commun, Vol. COM-28, pp. 84-95, January 1980 https://doi.org/10.1109/TCOM.1980.1094577
  2. W.H.Equitz, 'A new vector quantization clustering algorithm,' IEEE Trans. Acoust Speech and Signal Proc., Vol. 7, pp. 1568-1575, October 1989 https://doi.org/10.1109/29.35395
  3. I.Katsavounidis, C.C. Jay Kuo, and Z.Zhang, 'A new initialization technique for generalized Lloyd iteration,' IEEE Signal Processing Letters, Vol. 1, pp. 144-146, October 1994 https://doi.org/10.1109/97.329844
  4. H.A.Monawer, 'Image vector quantization using a modified LBG algorithm with approximated centroids,' Electronics Letters, Vol. 31, pp. 174 -175, February 1995 https://doi.org/10.1049/el:19950100
  5. D.Lee, S.Baek, and K.Sung, 'Modified K -means algorithm for vector quantizer design,' IEEE Sigml Processing Letters, Vol. 4, pp. 2~4, January 1997 https://doi.org/10.1109/97.551685
  6. S.Baek, B.Jeon, D.Lee and K.Sung, 'Fast clustering algorithm for vector quantization, 'Electronics Letters, Vol. 34, pp. 151 -152, January 1998 https://doi.org/10.1049/el:19980217
  7. P.Veprek and A.B.Bradley, 'An improved algorithm for vector quantizer design,' IEEE Signal Processing Letters, Vol. 7, pp. 250-252, September 2000 https://doi.org/10.1109/97.863147
  8. M.R.Anderberg, Cluster analysis for application, Academic, New York, 1973
  9. 박소희, 조제황, '초기 학습반복시 수렴영역을 벗어난 가중치에 K-meana 알고리즘,' 한국음향학회 추계학술발표대회 논문집, 제 20권, 제2(s)호, 143-146쪽, 2001년 11월