# US 2008/0186225 A1 · Markov Sequential Detector

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2008/0186225 A1
- Document ID: US20080186225A1
- Status: Published application
- Kind: A1
- Application number: 11/912,924
- Filed: 2006-04-18
- Published: 2008-08-07
- Earliest claimed: 2005-04-29
- Assignees: Thales
- Inventors: Billon Didier
- Web page: https://patentgrep.com/patent/US20080186225A1

## CPC

- G01S 15/66: Sonar tracking systems
- G01S 7/2923: {based on data belonging to a number of consecutive radar periods}
- G01S 13/66: Radar-tracking systems; Analogous systems
- G01S: RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES

## Abstract

The present invention relates to the general problem of detection by sonar or radar. It is concerned in particular with the initial detection of a signal of interest in a data stream corresponding to the signal received by such equipment. This initial detection is transmitted to the equipment charged with performing the tracking of the objects of interest that were the subject of a detection. The method considers that a signal of interest is characterized by a state model x(t) with values in a finite set {x1, . . . , xN}, to which there corresponds a finite set {ω1, . . . , ωN} of N values of an observation variable ω, characteristic of this signal. To detect the presence of a signal of interest the method simultaneously calculates the value of a detection criterion Λn(t) in two different ways. The first way, the detection function is calculated is as a function of the product of the probability Pn of the state xn, and of the ratio of the conditional probabilities p1 and p0 of the datum y(t, ωn) containing the signal, in the absence and in the presence of a signal. The second way the detection signal is calculated is as a function of the product of one of the N values of the criterion that were calculated at the previous instant t−1, of the probability of transition Pn,m between the state xm for which this value was calculated at the instant t−1 and the state xn considered, and of the ratio of the conditional probabilities p1 and p0 of the datum y(t, ωn) in the absence and in the presence of a signal. The larger of the two calculated values is assigned to the criterion of Λn(t) which is then compared with a detection threshold.

## Claims

### Claim 1

A continuous method of detecting a signal s(t) present in a data stream y(t, ω) dependent on a discrete time variable t and an observation variable ω, the signal being characterized by a state model x(t) with values in a finite set {x1, . . . , xN}, to which there corresponds a finite set {ω1, . . . , ωN} of N values of the observation variable ω, this method formulating for each instant t and for each state xn a detection criterion Λn(t),
wherein at each instant t the value of the criterion Λn(t) for a state xn is calculated simultaneously in two different ways:
as being a function of the product of the probability Pn of the state xn, and of the ratio of the conditional probabilities p1 and p0 of the datum y(t, ωn) in the absence and in the presence of a signal,
as being a function of the product of one of the N values of the criterion that were calculated at the previous instant t−1, of the probability of transition Pn,m between the state xm for which this value was calculated at the instant t−1 and the state xn considered, and of the ratio of the conditional probabilities p1 and p0 of the datum y(t, ωn) in the absence and in the presence of a signal,
the larger of the two calculated values being assigned to the criterion of Λn(t).

### Claim 2 (depends on 1)

The method as claimed in claim 1 further comprising
a step A of initialization where the N values of the detection criterion Λn at the instant t0−1 preceding the first instant of the data to be processed t0 are fixed at predetermined values,
a step B of calculating the N values of the criterion Λn at the current instant t according to the two ways and of determining the maximum value,
a step C of detecting by comparison with a threshold the N values of the criterion that were determined at phase B, the overshooting of the detection threshold by a value of the criterion determining the presence of a signal in the state xn corresponding to this value of the criterion, at the current instant, steps B and C being carried out in an iterative manner.

### Claim 3 (depends on 1)

The method as claimed in claim 1, in which the detection criterion Λn(t) is defined by the following relations:
Λ
n

(
t
0
-
1
)
=
0
,
Λ
n

(
t
)
=
max

{
F

(
K
2
·
p
1

(
y

(
t
,
ω
n
)
)
p
0

(
y

(
t
,
ω
n
)
)
·
max
1
≤
m
≤
N

{
P
n
,
m
·
Λ
m

(
t
-
1
)
}
)
,
F

(
K
1
·
p
1

(
y

(
t
,
ω
n
)
)
p
0

(
y

(
t
,
ω
n
)
)

P
n
)
}
where t0 is the first instant of the data to be processed, ωn is the locus in the observation space corresponding to the state xn, where p1 and p0 represent the conditional probability laws for the data y(t, ω) in the absence and in the presence of a signal, where F is an increasing function, and where K1 and K2 are two arbitrarily chosen factors.

### Claim 4 (depends on 3)

The method as claimed in claim 3, for which the factor K1 or the factor K2 is regulated so as to maintain a reinitialization rate for the recursive calculation of Λn that is determined by the condition:
F

(
K
1
·
p
1

(
y

(
t
,
ω
n
)
)
p
0

(
y

(
t
,
ω
n
)
)

P
n
)
>
F

(
K
2
·
p
1

(
y

(
t
,
ω
n
)
)
p
0

(
y

(
t
,
ω
n
)
)
·
max
1
≤
m
≤
N

{
P
n
,
m
·
Λ
m

(
t
-
1
)
}
)
,
near to a specified value R.

### Claim 5 (depends on 4)

The method as claimed in claim 4 for which R is equal to 0.5.

### Claim 6 (depends on 1)

The method as claimed in claim 1 applied to the initialization of sonar tracks on spectral lines on the basis of observation data y(t, ω) for which ω is a frequency datum.

### Claim 7 (depends on 1)

The method as claimed in claim 1 applied to the initialization of sonar tracks on spectral lines on the basis of observation data y(t, ω) for which ω is a data pair (frequency, bearing).

### Claim 8 (depends on 2)

The method as claimed in claim 2 applied to the initialization of sonar tracks on spectral lines on the basis of observation data y(t, ω) for which ω is a frequency datum.

### Claim 9 (depends on 4)

The method as claimed in claim 4 applied to the initialization of sonar tracks on spectral lines on the basis of observation data y(t, ω) for which ω is a frequency datum.

### Claim 10 (depends on 2)

The method as claimed in claim 2 applied to the initialization of sonar tracks on spectral lines on the basis of observation data y(t, ω) for which ω is a data pair (frequency, bearing).

### Claim 11 (depends on 4)

The method as claimed in claim 4 applied to the initialization of sonar tracks on spectral lines on the basis of observation data y(t, ω) for which ω is a data pair (frequency, bearing).

### Claim 12 (depends on 2)

The method as claimed in claim 2 in which the detection criterion Λn(t) is defined by the following relations:
Λ
n

(
t
0
-
1
)
=
0
,
Λ
n

(
t
)
=
max

{
F

(
K
2
·
p
1

(
y

(
t
,
ω
n
)
)
p
0

(
y

(
t
,
ω
n
)
)
·
max
1
≤
m
≤
N

{
P
n
,
m
·
Λ
m

(
t
-
1
)
}
)
,
F

(
K
1
·
p
1

(
y

(
t
,
ω
n
)
)
p
0

(
y

(
t
,
ω
n
)
)

P
n
)
}
where t0 is the first instant of the data to be processed, ωn is the locus in the observation space corresponding to the state xn, where p1 and p0 represent the conditional probability laws for the data y(t, ω) in the absence and in the presence of a signal, where F is an increasing function, and where K1 and K2 are two arbitrarily chosen factors.

### Claim 13 (depends on 12)

The method as claimed in claim 12 for which the factor K1 or the factor K2 is regulated so as to maintain a reinitialization rate for the recursive calculation of Λn that is determined by the condition:
F

(
K
1
·
p
1

(
y

(
t
,
ω
n
)
)
p
0

(
y

(
t
,
ω
n
)
)

P
n
)
>
F

(
K
2
·
p
1

(
y

(
t
,
ω
n
)
)
p
0

(
y

(
t
,
ω
n
)
)
·
max
1
≤
m
≤
N

{
P
n
,
m
·
Λ
m

(
t
-
1
)
}
)
,
near to a specified value R.

### Claim 14 (depends on 13)

The method as claimed in claim 13 for which R is equal to 0.5.
