© The Institution of Engineering and Technology
The authors previously considered the application of multilayer perceptrons (MLPs) to block coding at the sensor level of modern imaging systems, and have proposed analogue encoders with transistor-count complexity that is low enough to suit focal-plane implementation. In the paper, they extend the on-sensor block coding MLP study, to include a statistical analysis of the MLP sensitivity to implementation errors occuring in standard CMOS fabrication processes. Employing simple offset models, a comparison is made of the MLP with other block encoders based on full-search entropy-constrained vector quantisation (ECVQ) of the data, and it is verified that the MLPs are less sensitive over a wide range of rate-distortion compression points. By introducing a realistic linear model that takes into account sensitivity and the complexity performances for both systems, the authors verify that, for MLPs, the sensitivity becomes less dependent on the complexity as the expected quality loss is allowed to increase. Without setting a limit on the expected quality loss, the MLPs are consistently better than the ECVQs, both in terms of sensitivity and complexity for a precision equivalent to 6 bits.
References
-
-
1)
-
C. Mead
.
(1989)
Analog VLSI and neural systems.
-
2)
-
C. Alippi ,
M. Catelani ,
A. Fort ,
M. Mugnaini
.
SBT soft fault diagnosis in analog electronic circuits: a sensitivity-based approach by randomized algorithms.
IEEE Trans. Instrum. Meas.
,
5 ,
1116 -
1125
-
3)
-
Liñán, G.: `Diseño, de chips programables de señal mixta con bajo consumo de potencia para sistemas de vision en tiempo real', June 2002, PhD, University of Seville, Spain.
-
4)
-
X. Zeng ,
D.S. Yeung
.
Sensitivity analysis of multilayer perceptron to input and weight perturbations.
IEEE Trans. Neural Netw.
,
6 ,
1358 -
1366
-
5)
-
Liu, X., El Gamal, A.: `Simultaneous image formation and motion blur restoration via multiple capture', Proc. IEEE Int. Conf. Acoust. Speech Signal Process., Salt Lake City, UT, May 2001, p. III.1841–III.1844.
-
6)
-
E. Funatsu
.
Artificial retina large scale integration with on-sensor projection function for high-speed motion detection.
Opt. Eng.
,
11 ,
2709 -
2718
-
7)
-
P.A. Chou ,
T. Lookabaugh ,
R.M. Gray
.
Entropy-constrained vector quantization.
IEEE Trans. Acoust. Speech Signal Process.
,
1 ,
31 -
42
-
8)
-
S.-H. Oh ,
Y. Lee
.
Sensitivity analysis of single hidden-layer neural networks with threshold functions.
IEEE Trans. Neural Netw.
,
4 ,
1005 -
1007
-
9)
-
A.G. Andreou ,
K.A. Boahen ,
P.O. Pouliquen ,
A. Pavasoví ,
R.E. Jenkins ,
K. Strohbehn
.
Current-mode subthreshold MOS circuits for analog and VLSI neural systems.
IEEE Trans. Neural Netw.
,
2 ,
205 -
213
-
10)
-
G. Ferri ,
N.C. Guerrini
.
(2003)
Low-voltage low-power CMOS current conveyors.
-
11)
-
Y. Linde ,
A. Buzo ,
R.M. Gray
.
An algorithm for vector quantizer design.
IEEE Trans. Commun.
,
1 ,
84 -
95
-
12)
-
Gomes, J.G.R.C., Mello, M.J.C., Haas, H.L., Petraglia, A.: `New error sensitivity model for the analog hardware implementation of inner products', Proc. IEEE Int. Conf. Image Processing, 8–11 October 2006, Atlanta, GA, p. 3333–3336.
-
13)
-
Cauwenberghs, G., Waskiewicz, J.: `Analog VLSI cellular implementation of the boundary contour system', Proc. Neural Inf. Process. Syst. Conf., 30 November–5 December 1998, Denver, CO, p. 657–663.
-
14)
-
B. Scholkopf ,
A. Smola ,
K.R. Muller
.
Nonlinear component analysis as a kernel eigvalue problem.
Neural Comput.
,
1299 -
1319
-
15)
-
J.G.R.C. Gomes ,
S.K. Mitra
.
A comparative study of the complexities of neural network based focal-plane image compression schemes.
IEICE Trans. Fundam. Electron. Commun. Comput. Sci.
,
11 ,
1185 -
1196
-
16)
-
Malvar, H., Hallapuro, A., Karczewicz, M., Kerofsky, L.: `Low-complexity transform and quantization with 16-bit arithmetic for H.26L', Proc. IEEE Int. Conf. Image Process., September 2002, Rochester, NY, p. II.489–II.492.
-
17)
-
K. Bult ,
H. Waalinga
.
A class of analog CMOS circuits based on the square-law characteristic of an MOS transistor in saturation.
IEEE J. Solid-State Circuits
,
3 ,
357 -
365
-
18)
-
A.S. Sedra ,
G.W. Roberts ,
F. Gohh
.
The current conveyor: history, progress and new results.
IEE Proc. G, Circuits Devices Syst.
,
2 ,
78 -
87
-
19)
-
A. Gersho ,
R.M. Gray
.
(1991)
Vector quantization and signal compression.
-
20)
-
J. Madrenas ,
M. Verleysen ,
P. Thissen ,
J.L. Vox
.
A CMOS analog circuit for Gaussian functions.
IEEE Trans. Circuits Syst. II, Analog Digit. Signal Process.
,
1 ,
70 -
74
http://iet.metastore.ingenta.com/content/journals/10.1049/iet-cds_20050315
Related content
content/journals/10.1049/iet-cds_20050315
pub_keyword,iet_inspecKeyword,pub_concept
6
6