Page 37 - Electrical Engineering Dictionary

nal and null all interfering or jamming sig- adaptive FIR filter a finite impulse re-
nals. This is achieved by finding the correct sponse structure filter with adjustable coef-
weights (input excitations) to the elements ficients. The adjustment is controlled by an
comprising the array. adaptation algorithm such as the least mean
square (LMS) algorithm. They are used
adaptive coding a coding scheme that extensively in adaptive echo cancellers and
adapts itself in some fashion to its input or equalizers in communication systems.
output.
adaptive fuzzy system fuzzy inference
adaptive coding of transform coefficients system that can be trained on a data set
coding technique that is carried out by through the same learning techniques used
threshold sampling and exploiting masking for neural networks. Adaptive fuzzy systems
effects by variable quantization for differ- are able to incorporate domain knowledge
ent blocks. High detail blocks are coded about the target system given from human
with more quantization error than low de- experts in the form of fuzzy rules and numer-
tail blocks. This is done to take into ac- ical data in the form of input–output data sets
count masking and boundary distortion ef- of the system to be modeled. See also neural
fects. Transform coding becomes more at- network, fuzzy inference system.
tractive compared with DPCM when adap-
tive coding is used. The main drawback of adaptive intrafield predictors a tech-
adaptive transform coding is its sensitivity nique used for picture signal prediction based
to transmission bit errors due to synchro- on local properties of the signal or side infor-
nization problems at the decoder. See also mation if portions of local properties have
DPCM. not been transmitted. Intrafield methods re-
quire correlation with local information for
adaptive control a control methodology prediction purposes.
in which control parameters are continuously A common technique is to use a mea-
and automatically adjusted in response to sure of the directional correlation based on
be measured/estimated process variables to local pixels that have already been transmit-
achieve near-optimum system performance. ted. A predictor is chosen from a set to give
minimum prediction error. For example, the
adaptive critic learning technique where previous line or previous pixel can be used
the system learns to evaluate the actions of a for prediction, and the switching can then be
system (usually a controller) so as to provide done as follows:
a reinforcement signal that is an estimate of
the future value of the system’s current ac-
tion. ∩X = predictor for element X

Aif kB − Ck < kA − Bk
adaptive differential pulse code modula- = C otherwise
tion (ADPCM) a modulation scheme in
whichonlythedifferencebetweensuccessive An extension of this concept is called con-
signal samples is encoded for transmission, tour prediction where the direction of pixel A
and the quantization of the coding is adapted is determined by searching among E, B, C,
to the characteristics of the signal source. or G.

adaptive filtering a filtering strategy in adaptive logic network tree-structured
which filter coefficients or governing param- network whose leaves are the inputs and
eters evolve over time according to some up- whose root is the output. The first hidden
dating strategy to optimize some criterion. layer consists of linear threshold units and the

c
2000 by CRC Press LLC