Browse Prior Art Database

Statistically Based Adaptive Threshold

IP.com Disclosure Number: IPCOM000100406D
Original Publication Date: 1990-Apr-01
Included in the Prior Art Database: 2005-Mar-15
Document File: 2 page(s) / 73K

Publishing Venue

IBM

Related People

Christopher, RJ: AUTHOR [+3]

Abstract

Disclosed is a method which sets a signal threshold adaptively, based on statistical characteristics of the input signal. This method was originally intended for use in the control electronics for a touch- sensitive screen based on piezoelectric transducers. However, the method described can be generalized, and would be applicable in any digital signal-processing system that operates on signals which are bounded in time. An ability to adjust the threshold in response to changing conditions results in a more robust system. This ability can also relax or eliminate requirements on analog circuitry.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 52% of the total text.

Statistically Based Adaptive Threshold

       Disclosed is a method which sets a signal threshold
adaptively, based on statistical characteristics of the input signal.
 This method was originally intended for use in the control
electronics for a touch- sensitive screen based on piezoelectric
transducers.  However, the method described can be generalized, and
would be applicable in any digital signal-processing system that
operates on signals which are bounded in time.  An ability to adjust
the threshold in response to changing conditions results in a more
robust system.  This ability can also relax or eliminate requirements
on analog circuitry.

      A digital signal processing system samples its input signal(s)
at some predetermined rate.  The nature of the input is such that an
information-carrying signal is present sometimes, and not at others
(e.g., some transducer-based systems).  In the absence of an
information-carrying signal, the input sample is passed to a
statistics- deriving routine.  The statistics-deriving routine
digitally filters the sampled data and the square of the sampled
data.  The filtering operation performed is a first-order IIR filter.
Filter coefficients are set to approximate a weighted average over
some number of samples.  Once the two filtered values are computed, a
third quantity, the standard deviation, can be computed according to
the formula:

      The standard deviation provides a measure of the noise in the
sampled data.  As such, it can be useful for adjusting thresholds to
ignore ambient noise.  If the sample distribution is known, a
cumulative distribution curve can be used to relate probability of
false detection and standard deviation.  If the sample distribution
is unknown, then empirically-d...