## Dynamic Circular Cellular Networks for Adaptive Smoothing of Multi-Dimensional Signals (1998)

Venue: | In IEEE International Workshop on Cellular Neural Networks and their Applications |

Citations: | 1 - 1 self |

### BibTeX

@INPROCEEDINGS{Wiehler98dynamiccircular,

author = {K. Wiehler and R. Lembcke and R. -r. Grigat and J. Heers},

title = {Dynamic Circular Cellular Networks for Adaptive Smoothing of Multi-Dimensional Signals},

booktitle = {In IEEE International Workshop on Cellular Neural Networks and their Applications},

year = {1998},

pages = {313--318}

}

### OpenURL

### Abstract

In [10] a theoretical framework for locally-adaptive smoothing of multidimensional data was presented. Based on this framework we introduce a hardware efficient architecture suitable for mixed-mode VLSI implementation. Substantial shortcomings of analogue implementations are overcome by connecting all cells in a circular structure: i) influence of process parameter deviation ii) limited number of cells iii) input/output bottleneck. The connections between the analogue cells and the cells themselves are dynamically reconfigured. This results in a non-linear adaptive filter kernel which is shifted virtually over the signal vector of infinite length. A 1-d prototype with 32 cells has been fabricated using 0.8¯m CMOS-technology. The chip is fully functional with an overall error less than 1%; experimental results are presented in this paper.

### Citations

116 |
The CNN paradigm
- Chua, Roska
- 1993
(Show Context)
Citation Context ... ?N )sr; j = 0; : : : ; N \Gamma 1; j 6= i g ; i = 0; : : : ; N \Gamma 1 (1) ! \Delta ?N is the modulo N operation. Each cell C i denotes a corresponding dynamical system of a cellular neural network =-=[2]. Th-=-e embedding of these structures into the planar surface of a semiconductor is straight forward [4], see ��gure 3 for possible embeddings of 1-d and 2-d structures. For higher dimensions the demand... |

52 |
Characterization and modeling of mismatch in mos transistors for precision analog design
- Lakshmikumar, Hadaway, et al.
- 1986
(Show Context)
Citation Context ... results are often critical for the desired functionality of the chip. Additionally, the impact of the process parameter deviations increases with the down scaling of the standard CMOS processes used =-=[1, 7]. As an exam-=-ple the minimum possible standard deviation of the ooeset voltage of a transconductance ampli��er (��g. 1) is depicted in ��gure 2. The results were acquired by performing a multi-dimensio... |

51 |
Design of Analog Integrated Circuits and Systems
- Laker, Sansen
- 1994
(Show Context)
Citation Context ...ecessary to employ techniques which are insensitive to device mismatch. These techniques can be divided into passive and active working methods. The former ones are robust layout/schematic techniques =-=[6]-=- which have basically no eoeect on the functionality. The latter ones compensate the errors either by adding additional correcting 1 supported by the Deutsche Forschungsgemeinschaft, research programm... |

17 |
Design and learning with cellular neural networks
- Nossek
- 1994
(Show Context)
Citation Context ...al of 95% circuitry [12] or by reducing the input and/or output signal range [3]. A further method to achieve the best possible results is to trim the analogue coeOEcients by self learning algorithms =-=[8]-=-. Another important problem in designing massive parallel analogue hardware is the coupling to external devices (e. g. cameras, subsequent digital processing parts, etc.). This is especially true for ... |

11 |
CMOS resistive fuses for image smoothing and segmentation
- Yu, Decker, et al.
- 1992
(Show Context)
Citation Context ...;k ); where f(x) = ae 2 h x jxjsc ae 2 l x + ( 2 h \Gammas2 l ) c ae sgn(x) jxj ? c ae (8) It can be shown that this kind of relaxation network behaves similarly to non-linear resistive grid networks =-=[9]-=-. The mapping of the algorithm to the analogue hardware was straightforward using voltage controlled current sources (OTAs). The entities have been carefully designed and placed on the chip to minimis... |

11 |
Analog storage of adjustable synaptic weights
- Vittoz, Oguey, et al.
- 1991
(Show Context)
Citation Context ...ffset voltage [mV] Figure 2: Minimumpossible standard deviation of the ooeset voltage for a ��xed OTA area. The corresponding resolution in bits is computed for a con��dence interval of 95% ci=-=rcuitry [12]-=- or by reducing the input and/or output signal range [3]. A further method to achieve the best possible results is to trim the analogue coeOEcients by self learning algorithms [8]. Another important p... |

8 |
Mismatch Characterization of Submicron MOS Transistors
- Bastos, Steyaert, et al.
- 1997
(Show Context)
Citation Context ... results are often critical for the desired functionality of the chip. Additionally, the impact of the process parameter deviations increases with the down scaling of the standard CMOS processes used =-=[1, 7]. As an exam-=-ple the minimum possible standard deviation of the ooeset voltage of a transconductance ampli��er (��g. 1) is depicted in ��gure 2. The results were acquired by performing a multi-dimensio... |

6 |
Dynamics and Self-organization of Locally Coupled Neural Networks. Collection
- Thiran
- 1997
(Show Context)
Citation Context ... of the network, the eoeective length N 0 can be equal to N since there will be no reAEection from that boundary. 2.1 Dioeusion When processing signals with global dioeusion space-invariant templates =-=[10, 11], wh-=-ich guarantee a global asymptotic behaviour, a characteristic space constantscan be de��ned by \Deltax(k) ! c u(i) exp ` \Gamma jk \Gamma ij ' 8u(i) 2 S; i; k 2 D; S: sig. space; x(k): state of C ... |

2 |
A 0.8 ��m CMOS 2-D Programmable Mixed-Signal Focal-Plane Array Processor with On-Chip Binary Imaging and Instructions Storage
- Dominguez-Castro, Espejo, et al.
- 1997
(Show Context)
Citation Context ...iation of the ooeset voltage for a ��xed OTA area. The corresponding resolution in bits is computed for a con��dence interval of 95% circuitry [12] or by reducing the input and/or output signa=-=l range [3]-=-. A further method to achieve the best possible results is to trim the analogue coeOEcients by self learning algorithms [8]. Another important problem in designing massive parallel analogue hardware i... |

2 |
On globally asymptotically stable continuous-time CNNs for adaptive smoothing of multidimensional signals
- rr, Stiehl, et al.
- 1996
(Show Context)
Citation Context ...Str. 30, D 22527 Hamburg, Germany phone: +49 (0)40 5494-2576, -2581, -2453, fax: -2572 email: --heers, schnoerr, stiehl@informatik.uni-hamburg.de http://kogs-www.informatik.uni-hamburg.de Abstract In =-=[10] a t-=-heoretical framework for locally-adaptive smoothing of multidimensional data was presented. Based on this framework we introduce a hardware ef��cient architecture suitable for mixed-mode VLSI impl... |

1 |
Two-dimensional spatio-temporal dynamics of analog image processing neural networks
- Kobayashi, Matsumoto, et al.
- 1995
(Show Context)
Citation Context ... equilibrium. As a consequence, only a poor utilisation factor of the circuit can be achieved. 2 Dynamic circular cellular networks Circular networks are described by a closed n-dimensional structure =-=[5]. Fo-=-r the 1-d case the neighbourhoodsN r (i) of the cell C i within a radius r on a grid D is de��ned by: N r (i) = fC j j min (! i \Gamma j ?N ; !j \Gamma i ?N )sr; j = 0; : : : ; N \Gamma 1; j 6= i ... |

1 |
Auswirkungen von Bauteilst#rungen in der analogen Signalverarbeitung am Beispiel eines CMOS-Transkonduktanzverst#rkers f#r zellul#re neuronale Netze
- Wiehler, Klingspohr, et al.
- 1997
(Show Context)
Citation Context ...cquired by performing a multi-dimensional optimisation of the transistor dimensions (width and length for each MOST). The computations are based on a 0.7��m CMOS process for optimal layout conditi=-=ons [13]. It can-=- be seen from these results that for an OTA in strong inversion an area of at least 2100 ��m 2 (3100 ��m 2 for weak inversion) is required to reach an accuracy of 8 bits which is a common dyna... |