Tytuł artykułu
Autorzy
Identyfikatory
Warianty tytułu
Programowo sprzętowa implementacja w układach programowalnych
Języki publikacji
Abstrakty
Dynamic Time Warping procedure is widely used in pattern matching applications, such as speaker recognition systems. It allows to align elements of nonlinear time sequences, such as acoustic feature sequences of utterances that have different length. Software implementation of DTW algorithm requires a lot of computation power and thus it can occupy the most of available CPU time, leaving little resources to perform other necessary tasks. On the other hand, putting whole DTW into hardware is a complex and difficult process, mainly due to high memory requirements. Embedded memory blocks available in modern FPGAs cannot satisfy this requirements, thus external RAM chips have to be used. This paper proposes hardware-software solution with partitioning between embedded software application and hardware component. Altera FPGA device, with NiosII-based software system is used to implement the procedure.
Procedura dynamicznej normalizacji czasowej (DTW) jest powszechnie stosowanym narzędziem w problemach dopasowania wzorców, takich jak problem rozpoznawania mówcy. Procedura jest wymagająca obliczeniowo i ma regularną strukturę, natomiast wymaga dużych zasobów pamięci I skomplikowanych algorytmów dostępu do niej. W artykule przedstawiono programowo-sprzętową implementację algorytmu DTW, w której powtarzalne obliczenia realizowane są w sprzęcie, natomiast dostępem do pamięci zarządza mikroprocesor.
Rocznik
Tom
Strony
415--420
Opis fizyczny
Bibliogr. 11 poz., rys., tab.
Twórcy
autor
autor
autor
- Warsaw University of Technology, Institute of Telecommunications
Bibliografia
- [1] Putz – Leszczyńska J., Pacut A.: Hidden signature – a new solution for on-line verification using DTW, Proc. 42th Annual IEEE International Carnahan Conference on Security Technology, Oct. 2008, Prague, Czech Rep.. 2008, s. 162–166.
- [2] Boeva V., Kostadinova E.: A Hybrid DTW based Method for Integration Analysis of Time Series Data, icais, 2009 International Conference on Adaptive and Intelligent Systems, 2009, s. 49–54.
- [3] Sakoe H., S. Chiba S.: Dynamic Programming Optimization for Spoken Word Recognition, IEEE Transactions on Acoustics, Speech and Signal Processing, Vol. ASSP-26, s. 43–49.
- [4] Itakura, F. Minimum Prediction Residual Principle Applied to Speech Recognition. In IEEE Trans. Acoustics, Speech, and Signal Proc. vol. ASSP-23,1975, s. 52–72.
- [5] Kaczmarek A., Staworko M.: Application of dynamic time warping and cepstrograms to text-dependent speaker verification: the way to a better algorithm, In: Signal Processing Algorithms, Architectures, Arrangements and Applications, 24–26 September, Poznan, Poland, s. 169–174.
- [6] Salvador S., Chan P.: FastDTW: Toward Accurate Dynamic Time Warping in Linear Time and Space, 3rd Workshop on Mining Temporal and Sequential Data, 2004, s. 70–80.
- [7] Irwin M. J.: A Digit Pipelined Dynamic Time Warp Processor, IEEE Transactions on Acoustics, Speech and Signal Processing, Vol. 36, No. 9, September 1988.
- [8] Sundaresan V. K., Nichani S., Ranganathan N., Sanker R.: A VLSI Hardware Accelerator for Dynamic Time Warping, Proc. 11th IAPR International Conf. on Pattern Recognition, The Hague, The Netherlands, August/September 1992, Vol. IV, s. 27–30.
- [9] Stainhaouer N., Carayannis G.: New Parallel Implementations for DTW Algorithms, IEEE Transactions on Acoustic, Speech and Signal Processing, Vol. 38, No. 4, April 1990.
- [10] Charot F., Frison P., Quinton P.: Systolic Architectures for Connected Speech Recognition, IEEE Transactions on Acoustic, Speech and Signal Processing, Vol. ASSP-34, No. 4, August 1986.
- [11] Altera Corporation, http://www.altera.com, 2010.
Typ dokumentu
Bibliografia
Identyfikator YADDA
bwmeta1.element.baztech-article-BPG8-0033-0065