Sensor detection device, corresponding detection method and computer program

Patent metadata

Field	Value
Publication number	US-9235807-B2
Application number	US-201113884814-A
Country	US
Kind code	B2
Filing date	Nov 8, 2011
Priority date	Nov 10, 2010
Publication date	Jan 12, 2016
Grant date	Jan 12, 2016

Abstract

A device detecting at least one scenario from predetermined scenarios wherein a physical system observed by at least one sensor is likely to be found, including: at least one sensor providing a physical system observation data sequence; a storage storing at least one statistical model associating possible observation data values with the predetermined scenarios; a computer programmed to select at least one of the scenarios based on the observation data sequence and at least one statistical model. The storage includes plural statistical models broken down into plural ordered levels between an input level, wherein each statistical model associates possible values of at least a portion of the observation data with intermediate states for the statistical model, and an output level, wherein each statistical model associating possible values of at least a portion of the intermediate states of a lower level with at least a portion of the predetermined scenarios.

First claim

The invention claimed is:

1. A device detecting at least one scenario from a set of predetermined scenarios wherein a physical system observed by at least one sensor is likely to be found, on the basis of physical system observation data provided by the sensor, comprising:
at least one sensor providing a physical system observation data sequence;
a memory for storing at least one statistical model associating possible observation data values with the predetermined scenarios;
circuitry, connected to the sensor and the memory, programmed to select at least one of the scenarios, from the plurality of predetermined scenarios, based on the observation data sequence and the at least one statistical model;
wherein the memory stores a plurality of statistical models broken down into a plurality of ordered levels between:
a first level, or input level, wherein each statistical model associates possible values of at least a portion of the observation data with intermediate states that are specific to said each statistical model; and
a final level, or output level, wherein each statistical model associates possible values of at least a portion of the intermediate states of at least one statistical model of a lower level with at least a portion of the predetermined scenarios;

wherein the circuitry is further programmed to:
determine a sequence of standardized measurement values of at least a portion of the intermediate states based on at least one statistical model specifically defined to determine that portion of the intermediate states and on values received by said at least one statistical model, wherein measurement values of an intermediate state suitable for being determined by a given statistical model are standardized when the sum of the measurement values of all the intermediate states suitable for being determinable by the given statistical model and at the same time is equal to a predetermined time-independent constant; and
select the at least one scenario, on the basis of the sequence of standardized measurement values and at least one output level statistical model,

wherein each statistical model of a level other than the output level has intermediate states from each of which the output is at least part of the sequence of standardized measurement values by imposing that the sum at each time of the intermediate state measurement values of each statistical model of a level other than the output level is equal to a time-independent constant, and the output level determines directly from the outputted sequences of standardized measurement values either one of the scenarios which is selected among the predetermined scenarios or a sequence of scenarios from among the predetermined scenarios.

2. A detection device according to claim 1 , wherein the circuitry is further programmed to determine the sequence of standardized measurement values of all the intermediate states based on all the statistical models defined to determine the intermediate states and on values received by the statistical models.

3. A detection device according to claim 1 , wherein the standardized measurement values are probability values.

4. A detection device according to claim 3 , wherein at least one statistical model of a level other than the output level is a hidden Markov model, and wherein the circuitry is further programmed to determine a sequence of probability values of at least a portion of the hidden states thereof iteratively on the basis of a data sequence provided at the input of the Markov model according to a recurrence relation between successive values of the probability value sequence, the recurrence relation involving a single value, dependent on the recurrence index, of the input data sequence provided.

5. A detection device according to claim 4 , wherein the recurrence relation takes a form of:
I NITIALIZATION :

p
~

⁡

(

X
0

=

i
|

Y
0

)

=

p
⁡

(

X
0

=
i

)

⁢

p
⁡

(

Y
0

|

X
0

=
i

)

⁢

⁢
and

p
⁡

(

X
0

=

i
|

Y
0

)

=

p
~

⁡

(

X
0

=

i
|

Y
0

)

∑
j

⁢

p
~

⁡

(

X
0

=

j
|

Y
0

)

Assignees

• Jallon Pierre
• Commissariat Energie Atomique

Inventors

• Jallon Pierre

Classifications

• G06N5/01

  Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound · CPC title

• G06N5/04

  Inference or reasoning models · CPC title

• G06F19/345

  Physics · mapped topic

• G06N99/005

  Physics · mapped topic

• G06N5/02 Primary

  Knowledge representation; Symbolic representation · CPC title