A Charge-Based CMOS Parallel Analog Vector Quantizer

We present an analog VLSI chip for parallel analog vector quantization. The MOSIS 2.0 µm double-poly CMOS Tiny chip contains an array of 16 × 16 charge-based distance estimation cells, implementing a mean absolute difference (MAD) metric operating on a 16-input analog vector field and 16 analog template vectors. The distance cell including dynamic template storage measures 60 × 78 µm2. Additionally, the chip features a winner-take-all (WTA) output circuit of linear complexity, with global positive feedback for fast and decisive settling of a single winner output. Experimental results on the complete 16 × 16 VQ system demonstrate correct operation with 34 dB analog input dynamic range and 3 µsec cycle time at 0.7 mW power dissipation.

[1]  Teuvo Kohonen,et al.  Self-Organization and Associative Memory , 1988 .

[2]  John Lazzaro,et al.  Winner-Take-All Networks of O(N) Complexity , 1988, NIPS.

[3]  Teuvo Kohonen,et al.  Self-organization and associative memory: 3rd edition , 1989 .

[4]  Carver Mead,et al.  Analog VLSI and neural systems , 1989 .

[5]  Andreas G. Andreou,et al.  Current-mode subthreshold MOS circuits for analog VLSI neural systems , 1991, IEEE Trans. Neural Networks.

[6]  Allen Gersho,et al.  Vector quantization and signal compression , 1991, The Kluwer international series in engineering and computer science.

[7]  Yuping He,et al.  A charge-based on-chip adaptation Kohonen neural network , 1993, IEEE Trans. Neural Networks.

[8]  Rama Chellappa,et al.  VLSI neuroprocessors for video motion detection , 1993, IEEE Trans. Neural Networks.

[9]  Raoul Tawel.  Real-time focal-plane image compression , 1993, [Proceedings] DCC `93: Data Compression Conference.