WO2006015888A1 - Method for locating an impact on a surface and device for implementing such a method - Google Patents

Abstract

Method for locating an impact on a surface (9), in which acoustic sensors (6) pick up acoustic signals ski (t) generated by the impact and the impact is located by calculating, for a number of reference points of index j, a validation parameter representative of a function: PRODkji1i2..i2P (ω)=øSkil (ω) ØRjil (ω) * øSki2 (ω) * øRji2 (ω) ...ØSki2p (ω)* ØRji2p (ω) where: ØSki (ω) and ØRji (ω)* are complex phases of Ski (ω) and of Rji (ω) , for i = i1, i2, ... , i2p, indices denoting sensors, Ski(ω) and Rji(ω) being the Fourier transform of ski (t) and rji (t) , rji (t) being a reference signal corresponding to the sensor i for an impact at the reference point j, p being a non-zero integer no greater than NSENS/2.

Description

Method for locating an impact on a surface and device for implementing such a method

The present invention relates to the methods for locating an impact on a surface and to the devices for implementing such methods.

More particularly, the invention relates to a method in which an impact is located on a surface belonging to an object provided with at least N_SENS acoustic sensors (the object forming the acoustic interface can be made of a single piece or of several items, joined together or at least in mutual contact) , N_SENS being a natural integer at least equal to 2, method in which:

- N_SENS signals S_ki(t) from acoustic waves generated in the object by said impact are respectively sensed by the sensors, i being an index between 1 and NSENS which denotes the corresponding sensor, - for each reference point of index j, at least one validation parameter representative of at least one intercorrelation of complex phases representative of said signals s_ki(t) and of reference signals r-ji(t) is calculated, each reference signal r-ji(t) corresponding to the signal that would be received by the sensor i in case of impact at a reference point j out of N_REF reference points belonging to said surface, N_REF being a natural integer at least equal to 1 and j being an index between 1 and N_REP, - and the impact is located by determining at least one reference point as close as possible to the point of impact, by applying at least one validation criterion to the validation parameter.

Document WO-A-03/107261 describes an example of such a method which already gives total satisfaction. The main object of the present invention is to further refine this method, in particular to obtain an even more stable and more reliable signal processing. To this end, according to the invention, a method of the type concerned is characterized in that said validation parameter is representative of at least one intercorrelation: PRODyV₂" "¹^ (ω) =φS_idl (ω) 0Rj₁₁ (ω) ^*φS_]ύ2 (ω) ^*0R_ji2 (ω) . . . ψS_U2p (ω) >R_ji2p (ω) where :

- φSki(ω) is the complex phase of S_ki(ω) , for i = ii, i₂,

... , l2p,

- φRji(ω) is the complex phase of R_ji (ω) , for i = i_1; i_2/ ... , i_2p,

- ^* denotes the conjugate operator, applied to øS_ki (ω) for i = i_2m and to øRji (ω) for i = i_2m-i, tn being an integer between 1 and p;

- S_ki(ω) is the Fourier transform of S_ki(t) , - Rji(ω) is the Fourier transform of r_ji(t) ,

- p is a non-zero natural integer less than or equal to

Ns_ENs/2 ,

- I₁, i₂, ... i_2p are 2p indices denoting 2p sensors, each between 1 and NSEN_S•

With these provisions, an impact positioning method is obtained which is particularly reliable, in particular because the abovementioned intercorrelation is independent of the type of impact and of the response from the sensors.

In various embodiments of the invention, there may, if necessary, be used also one and/or other of the following provisions: - the method includes a step for calculating the intercorrelation PRODkjW"Sp(ω) , then a step for the inverse Fourier transform of this intercorrelation to obtain a time function prod_kj ¹i¹ ₂"¹ _2p(t) from which said validation parameter is then calculated; - the intercorrelation

(ω) is standardized before proceeding with the inverse Fourier transform;

- for each reference point j , a resemblance function V_kj (t) is calculated, chosen from:

. prod_kjV2"SpCt) , . and a linear combination of a number of functions prodkjW

corresponding to a number of subsets of 2p sensors out of N_SE_NS;

- NSENS is an even number and V_kj (t) is proportional to prod_kj¹i¹2"'¹2p(t) , where p = N_SENS/2;

- NSENS = 3 and, for each reference point j , a resemblance function Vkj (t) is calculated, chosen from:

. V_kj(t) = a3. [prod_kj ¹²(t) + prod_kj ²³ (t) ] , . and V_kj(t) = b3. [prod_kj ¹²(t) + prod_kj ²³(t) + prod_kj ¹³ (t)] , a3 and b3 being constants;

- a3 = 1/2 and b3 = 1/3;

- NSENS = 4 and, for each reference point j , a resemblance function V_kj (t) is calculated, chosen from:

. V_kj(t) = a4.prod_kj ¹²³⁴(t), . V_kj(t) = b4. [prod_kj ¹² (t) + prod_kj ³⁴ (t) ] , . V_kj(t) = c4. [prod_kj ¹² (t) + prod_kj ²³(t) + prod_kj ³⁴(t) + prod_kj ¹⁴(t)] , a4 , b4 and c4 being constants;

- a4 = 1, b4 = 1/2 and c4 = 1/4;

- the validation parameter is chosen from: . MAXIMUMoj = |V_kj (t=0) I , . MAXIMUM₁J = Max (|V_kj (t) | ) ,

Max(MAXIMUM_0j)

. CONTRASTij=- ....

(EJMAXIMUMiJ-MaX(MAXIMUMIj))Z(NREF-I)

. ENERGY = ∑i (∑_t [s_ki(t)]²);

- only V_kj (t) is calculated for the time t = 0 roughly corresponding to the impact, as being the real part of PROD_kj ¹i¹ ₂"¹ ₂p(ω) , and MAXIMlM₀J is used as validation parameter;

- at least one validation criterion is used, chosen from the following criteria:

. CONTRAST^ > THRESHOLD₁, with THRESHOLD₁ > 1, . CONTRAST _xj > THRESHOLD₁ and

MAXIMUMoj / MAXIMUM₁J¹*⁰¹⁵³¹³ > THRESHOLD₂, where THRESHOLD₂ > 1 and MAXIMUM₁J^1101313 corresponds to the parameter MAXIMUM₁J of the signals processed previously and not having resulted in validation, . MAXIMUMOJ > THRESHOLD₃, with THRESHOLD₃ > 0 and

MAXIMUMoj / MAXIMUMoj^N0ISE > THRESHOLD₄ with THRESHOLD₄

> 1, where MAXIMUMoj^NOISE corresponds to the parameter MAXIMUMoj of the signals processed previously and not having resulted in validation, . MAXIMUMoj / Average (MAXIMUMoj¹*⁰¹^) > THRESHOLD₅, with

THRESHOLD₅ > 1,

. ENERGY / ENERGY^N0ISE > THRESHOLD₆, where THRESHOLD₆ > 1 and ENERGY^N0ISE corresponds to the parameter ENERGY of the signals processed previously and not having resulted in validation;

- the reference signals are predetermined theoretically; - the reference signals are used with said object without training phase;

- the reference signals are previously learned on a reference device identical to said object, then are used with said object without training phase.

Moreover, the invention also relates to a device specially adapted to implement a method as defined above.

Other characteristics and advantages of the invention will become apparent from the following description of one of its embodiments, given by way of non-limiting example, in light of the appended drawings.

In the drawings:

- Figure 1 is a perspective diagrammatic view showing an exemplary device comprising an acoustic interface designed to implement a method according to an embodiment of the invention, - and Figure 2 is a block diagram of the device of Figure 1.

In the different figures, the same references denote identical or similar elements. Figure 1 represents a device 1 designed to implement the present invention, which includes, for example:

- a microcomputer central processing unit 2, - a screen 3 linked to the central processing unit 2,

- and a solid object 5 on which the central processing unit 2 can identify the position of an impact, as will be explained below.

The object 5 can be of any type (table, shelf, window pane, wall, door, window, computer screen, display panel, interactive terminal, toy, vehicle dashboard, seat back, floor, vehicle shock absorber, etc.) in which acoustic waves (in particular Lamb waves) can be made to propagate by generating impacts on its surface 9, as will be explained below.

At least two acoustic sensors 6 are fixed to the object 5 and are linked, for example, to microphone inputs 7 of the central processing unit 2, via cables 8 or by any other transmission means (radio, infrared or other) , so that said acoustic waves can be captured and transmitted to the central processing unit 2. The term N_SENS will hereinafter be used to denote the number of sensors and each sensor will be identified by an index i from 1 to N_SENS-

The acoustic sensors 6 can be, for example, piezo¬ electric sensors, or others (for example, capacitive sensors, magnetostrictive sensors, electromagnetic sensors, acoustic velocimeters, optical sensors [laser interferometers, laser vibrometers, etc.] , etc.) . They can be designed to measure, for example, the amplitudes of the movements due to the propagation of the acoustic waves in the object 5, or even the speed or the acceleration of such movements, or there may even be a pressure sensor measuring the pressure variations due to the propagation of the acoustic waves in the object 5. To enable the central processing unit 2 to locate an impact on the surface 9, the signals received by the sensors i when an impact is generated at a certain number N_REF of reference points 10 (each identified by an index j from 1 to N_REF) on the surface 9 are first of all determined. In the example represented in Figure 1, the surface 9 forms an acoustic interface in keyboard form, and the areas forming the reference points 10 can, if necessary, be marked by markings delimiting these areas and the information associated with them.

To this end, the first stage of the method can be a training step during which impacts are generated at reference points j of the surface 9.

These impacts can be generated, for example, by successively exciting the reference points j with any tool (including a part of the human body, such as a nail) , advantageously a tool, the contact surface of which remains constant in time. The force of the impact is, for example, perpendicular to the surface 9 or oriented in a constant direction.

On each impact, the impulse responses are detected by the sensors 6 and stored by the central processing unit 2 to form a bank of so-called reference signals, denoted r-_ji(t) (reference signal detected by the sensor numbered i for an excitation of the reference point numbered j) . There are N_SENS ^• N_REF of these reference signals.

As a variant, the reference signals are predetermined theoretically, and, where appropriate, then used with said object 5 without training phase.

According to another variant, the reference signals can be learned previously on a reference device identical to said object 5, then, where appropriate, are used with said object 5 without learning phase.

The reference signals r-ji(t) can be expressed as follows:

r-ji(t) = e-_j(t) * hji(t) * mi(t) (1)

With: βj (t) time function of the excitation force, hji(t) impulse response (Green's function) for a force applied to the point j and detected by the sensor i, πii(t) impulse response of the sensor i,

* symbol representing the time convolution operator.

By switching to the frequency domain, the equation (1) becomes:

R_ji(ω) = E_j (ω) Hji(ω) M₁(CU) (2)

With:

E_j (ω) : Fourier transform of e_j (t) H_ji (ω) : Fourier transform of hji(t) Mi (ω) : Fourier transform of mi(t)

After the training step, the device 1 is used to locate any impact at a point numbered k of the surface 9. The N_SENS signals newly detected by the sensors i, denoted s_ki(t) are then compared with the reference signals r_ji(t) stored previously, so as to- determine whether the point of impact k corresponds to a known reference point j .

To this end, S_k1 (ω) , the N_SENS Fourier transforms of the signals S_ki(t) are determined first of all, then the products M_kji of the exponential phases of the new signals S_ki(ω) with the phases of the signals Rji(ω)^* are then determined (the index i corresponds to the number of the sensor and the ^* sign indicates the conjugate complex) . To simplify the notation, the prefix φ will be used below to indicate the exponential phase of a complex variable, for example: φS_ki(ω) is the exponential phase of Ski(ω) and φR_ji(ω) is the exponential phase of R_ji(ω) .

Each product I%ji can therefore be written:

M_kji(ω) = φSki(ω) φRj± (ω) ^* (3)

If S_ki(ω) and Rji(ω) are broken down according to the equation (2) :

S_ki(ω) = Ek(ω) Hid(ω) M₁(ω) and R_ji(ω) = Ej (ω) Hji(ω) Mi(ω)

Hence:

iy^ji(ω) = φE_k(ω) φHki(ω) φMi(ω) φEj (co)^* φHji(ω)^* φMi(ω)^* (3')

One or more correlation products

are then calculated, respectively by correlation of an even number 2p of signals M_kji(ω) originating from 2p sensors ii, X₂, ... i2p out of the N_SENS sensors (p is an integer from 1 to Ns_ENs/2) :

PROD_kj ⁱiⁱ ₂"'ⁱ ₂p(ω) = M_kjii(ω)M_kji₂(ω)^*..M_kj_i2p(ω)^* (4)

It will be noted that the conjugation operator * is applied to M_kji (ω)φS_ki (ω) with i = i_2m/ iti being an integer between 1 and p.

If the formula (4) is developed, given that:

φE_k(ω)φE_k(ω) ^* φEj (ω)φEj (ω)^* = 1 and φM₁(ω)φM₁(ω)^* φM₂(ω)φM₂(ω)^* = 1,

we obtain: PRODkJV₂ -Sp (ω) = 0Hj₁₁ (ω) ^* øH_kil (ω) φH_{j i2} (ω) ^* φH_ki2 (ω) . . . øH_{j 2} (co) ^* φH_k2p (ω) (4 ¹ )

It can be noted that PRODkjWVp(co) does not depend on the type of excitation or on the response of the sensors, which makes it an extremely interesting variable in comparing the signal received from the impact at the point k with the reference signals of the signal bank, in order to determine the position of the point k.

The PRODkjWVp(Co) value or values can be standardized, for example as follows:

(_Npts /

' PROD_kj'Λ-Vp(ω) (5)

Where:

PRODkjSS"Sp^N(ω) is the standardized value of PROD_kjV₂"S_p(OJ) ,

N_pts = is the duration of the time signals Ski(t) and r-ji(t) picked up by the sensors 6 and stored by the central processing unit 2, that is, the number of points of each of these signals after sampling and digitization,

I¹⁵³V₂--Sp = ∑ω I PRODkΛYΛpCω) I (6)

^{is the} integral of | PROD_kjV₂"Sp(Co) | over the frequencies) .

It will be noted that the standardized value can, if necessary, be calculated directly, without separately determining PROD_kj ⁱ ₁ ¹ ₂"¹ _2p(ω) .

The next step is to return to the time domain by calculating the inverse Fourier transform of PRODISV'S_p(CO) , or prod_kj ¹ ₁S--S_p(t) . To determine whether the point k corresponds to one of the reference points J₀, resemblance functions V_kj (t) are used, which can, depending on case, be equal (or more generally proportional) respectively to: _prod_kjⁱiⁱ ₂"ⁱ ₂p(t) ,

- or a linear combination (in particular an average) of a number of functions prod_kj¹i¹ ₂"'^X2_P(t) corresponding to a number of subsets of 2p sensors.

As an example, for N_SENS = 2 sensors, the resemblance functions V_ki(t) are equal (or more generally proportional) respectively to prod_kj¹²(t) (ii=l, i₂=2 and

P=D.

For N_SENs = 3 sensors, the resemblance functions V_ki(t) can be chosen to be equal (or more generally proportional) respectively to:

- either 1/2 • [prod_kj ¹² (t) + prod_kj ²³ (t) ] (respectively ii=l, i₂=2 and p=l for prod_kj¹²(t) , ii=2, i₂=3 and p=l for prod_kj ²³ (t) ) ,

- or 1/3- [prod_kj ¹² (t) + prod_kj ²³(t) + prod_kj ¹³ (t) ] (respectively ii=l, i₂=2 and p=l for prod_kj¹²(t), ii=2, i₂=3 and p=l for prod_kj ²³(t), ii=l, i₂=3 and p=l for prodkj¹³ (t) ) .

For N_SENS = 4 sensors, the resemblance functions V_ki(t) can be chosen to be equal (or more generally proportional) respectively to: - prod_kj ¹²³⁴(t)

and p=2) ,

- or 1/2 • [prod_kj ¹² (t) + prod_kj ³⁴(t)] (respectively ii=l, i₂=2 and p=l for prod_kj ¹² (t) ,

and p=l for prod_kj ³⁴ (t)) ,

- or 1/4- [prod_kj ¹² (t) + prod_kj ²³ (t) + prod_kj ³⁴(t) + prod_kj ¹⁴ (t)] (respectively

and p=l for prod_kj ¹² (t) , ii=2, i₂=3 and p=l for prod_kj²³ (t) , ii=3 , i₂=4 and p=l for prod_kj ³⁴ (t) ,. ii=l, i₂=4 and p=l for prodkj¹⁴ (t)) . For a number of sensors greater than 4, the procedure is similar to the cases described above, as explained previously, by calculating the functions V_kj (t) as being equal (or more generally proportional) respectively to: - prod_kj¹ _I ¹ _Z ^■•i _2P(t) (in particular if N_SENS = 2p, and ii=l,

/

- or a linear combination of a number of functions prodkjW^-12p(t) corresponding to a number of subsets of 2p sensors out of N_SENS: in particular, Vkj (t) can be equal (or more generally proportional) to 1/n times the sum of n functions

corresponding to n different subgroups of 2p sensors out of N_SENS/ each sensor i preferably being included at least once in these subgroups.

In these different examples, each resemblance function V_kj (t) is between -1 and 1.

With the resemblance functions determined, these functions are used to calculate one or more validation parameters.

As an example, one or more of the following validation parameters can be calculated: - MAXIMUMoj = |V_kj (t=0) |

- MAXIMUM₁J = Max ( | V_kj (t) | ) MaX(MAXIMUM₀.)

- CONfRAST₁J =^"- — ; ^: — —7- —

(∑j MAXIMUMij - Max (MAXIMUMIJ) ) / (NREF ^~ 1 )

- ENERGY = ∑i ( ∑_t [ S_ki ( t ) ] ² ) .

Thus, it is possible to validate an impact located at the point k and confirm that it is located at a reference point jθ, for example, if it satisfies at least one of the following groups of criteria:

1) Group 1:

The impact k is considered located at the point jθ if: CONTRAST₁J₀ > CONTRAST₁J for j different from J₀ and CONTRAST₁JO > THRESHOLD₁, with THRESHOLD₁ > 1, for example THRESHOLD₁ = 2.

2) Group 2:

The impact k is considered located at the point jO if

CONTRAST₁JO > CONTRAST₁J for j different from J₀ and

CONTRAST₁JO > THRESHOLD₁ and

MAXIMUMojo / MAXIMUM_1Jo^NOISE > THRESHOLD₂ where THRESHOLD₂ > 1, for example THRESHOLD₂ = 2, and

corresponds to the parameter MAXIMUM_1J0

(averaged or not) of the signals processed previously and not having resulted in validation. This criterion avoids validating noise or random successive impacts on the object 5.

3) Group 3:

The impact k is considered located at the point jO if MAXIMUMojo > MAXIMUMoj for j different from J₀ and

MAXIMUMojo > THRESHOLD₃, with THRESHOLD₃ > 0, for example THRESHOLD₃ = 0.5, and

MAXIMUMojo / MAXIMUMojo^NOISE > THRESHOLD₄ with THRESHOLD₄ > 1, for example THRESHOLD₄ = 2.

MAXIMUMojo^NOISE corresponds to the parameter MAXIMUMojo (averaged or not) of the signals processed previously and not having resulted in validation.

Furthermore, for refinement purposes, the following criterion can be added:

MAXIMUMOJO / Average(MAXIMUMoj₀ ^NOISE) > THRESHOLD₅ (7) with THRESHOLD₅ > 1, for example THRESHOLD₅ = 4 In addition to these criteria, the following energy- criterion can also be added:

ENERGY / ENERGY^N0ISE > THRESHOLD₆ (8) where THRESHOLD₆ > 1, for example THRESHOLD₆ = 2 and ENERGY^N0ISE corresponds to the parameter ENERGY -(averaged or not) of the signals processed previously and not having resulted in validation.

As an example, it is therefore possible to validate a point of impact k and determine that it corresponds to a reference point jθ by using one of the following combinations of criteria and groups of criteria:

- (group 1) ,

- (group 2) , - (group 3) ,

- (group 1) or (group 2) ,

- (group 2) or (group 3) ,

- (group 1) or (group 3) ,

- (group 1) or (group 2) or (group 3) , - (group 1) and (7) ,

- (group 2) and (7) ,

- (group 3) and (7) ,

- [(group 1) or (group 2)] and (7),

- [(group 2) or (group 3)] and (7), - [(group 1) or (group 3)] and (7),

- [(group 1) or (group 2) or (group 3)] and (7), (-g-roup 1)—and-(8), - — — —

- (group 2) and (8) ,

- (group 3) and (8) , - [(group 1) or (group 2)] and (8),

- [(group 2) or (group 3)] and (8),

- [(group 1) or (group 3)] and (8),

- [(group 1) or (group 2) or (group 3)] and (8),

- (group 1) and (7) and (8) , - (group 2) and (7) and (8) ,

- (group 3) and (7) and (8),

- [(group 1) or (group 2)] and (7) and (8),

- [(group 2) or (group 3)] and (7) and (8),

- [(group 1) or (group 3)] and (7) and (8), - [(group 1) or (group 2) or (group 3)] and (7) and (8) .

The central processing unit 2 can thus locate the point of impact k on the surface 9 of the object 5. The determination of this point of impact may, if necessary, be the only information sought by the central processing unit, or it may even, if necessary, be used by said central processing unit 2 to deduce other information from it, for example a predetermined item of information assigned to a location on the surface 9 (the surface 9 can thus constitute an acoustic keypad) . Said information assigned to a location on the surface 9 can be predetermined information assigned in advance to said location, or even information determined dynamically on each new impact on the surface 9, according to impacts previously detected.

It will be noted that, when the parameter MAXIMUM_Oj is used to validate the impacts, for example when the

(group 3) is used as the validation criterion, it is possible to calculate values approximating to MAXIMUM_Oj very simply and quickly. In practice, when k is equal to j , terms of type øH_jX(ω)^* φH_ki(ω) of the equation (4¹) are purely real. Now, simply summing the real part of the product PROD_kj¹i¹ ₂"'^:L ₂p(ω) amounts to calculating the inverse Fourier transform of the equation (4') at the time t=0.

According to this variant, the real part of each product PROD_kj ¹i¹ ₂-"¹ ₂p (ω) is therefore calculated, which gives a value approximating to the value of prod_kj ⁱi^x ₂"ⁱ _2p(t) at t = 0. From this, the value of V_kj(0) is then deduced as explained previously, which gives the value of the parameter MAXIMUM_Oj, then the or each required validation criterion, for example the abovementioned group 3, is applied. With this variant, the calculation load is far less and continuous monitoring is much more easily, possible, even with a large number of reference points.

Moreover, the point of impact k on the surface 9 can be __PQSitioned,_ even when it .is .not _on__one_of_ .the reference points, by interpolation, as explained in the abovementioned document WO-A-03/107261.

Moreover, the reference signals can theoretically be modelled and applied to real objects, the acoustic characteristics of which are the same as those of the object concerned and used in the modelling.

The reference signals learned or theoretically modelled can be applied without training phase to objects having acoustic characteristics identical to those of the object used for training the reference signals or considered and used in the modelling.

Claims

1. Method in which an impact is located on a surface (9) belonging to an object (5) provided with at least N_SENS acoustic sensors (6) , N_SENS being a .natural integer at least equal to 2, . in which method:

- NSENS signals Ski(t) from acoustic waves generated in the object by said impact are respectively sensed by the sensors (6) , i being an index between 1 and N_SENS which denotes the corresponding sensor (6) ,

- for each reference point (10) of index j, at least one validation parameter representative of at least one intercorrelation of complex phases representative of said signals S_ki(t) and of reference signals r-ji(t) is calculated, each reference signal r_ji(t) corresponding to the signal that would be received by the sensor i (6) in case of impact at a reference point j (10) out of N_REF reference points (10) belonging to said surface (9) , N_REF being a natural integer at least equal to 1 and j being an index between 1 and NREF, - and the impact is located by determining at least one reference point (10) as close as possible to the point of impact, by applying at least one validation criterion to the validation parameter, characterized in that said validation parameter is representative of at least one intercorrelation:

where:

- φS_ki(ω) is the complex phase of Ski(ω) , for i =

- φR_ji(ω) is the complex phase of Rji(ω) , for i =

- ^* denotes the conjugate operator, applied to φS_ki(ω) for i = i_2m and to φRji(ω) for i = i_2m-i/ ™ being an integer between 1 and p;

- S_ki(ω) is the Fourier transform of s_ki(t),

- Rji(ω) is the Fourier transform of r_ji(t),

- p is a non-zero natural integer less than or equal to N_SENS/2,

- ix, i₂, ... i2_P are 2p indices denoting 2p sensors (6) , each between 1 and N_SENS■

2. Method according to Claim 1, including a step for calculating the intercorrelation

, then a step for the inverse Fourier transform of this intercorrelation to obtain a time function prodkjW^>:L2p(t) from which said validation parameter is then calculated.

3. Method according to Claim 2, wherein the intercorrelation PRODkjW^-:L2p(ω) is standardized before proceeding with the inverse Fourier transform.

4. Method according to Claim 2 or Claim 3, wherein for each reference point j (10) , a resemblance function V_kj (t) is calculated, chosen from:

- prod_kjV₂-'Sp(t) , - and a linear combination of a number of functions

corresponding to a number of subsets of 2p sensors (6) out of N_SENS-

5. Method according to Claim 4, wherein N_SENS is an even number and V_kj (t) is proportional to prod_kjV₂-"Sp(U _/ where p = N_SENS/2.

6. Method according to Claim 5, wherein N_SENS = 2 and V_kj (t) is proportional to prodkj¹² (t) •

7. Method according to Claim 4, wherein NS_ENS = 3 and, for each reference point j (10) , a resemblance function V_kj(t) is calculated, chosen from:

- V_kj(t) = a3. [prodkj¹²(t) + prod_kj ²³ (t) ] , - and V_kj(t) = b3. [prod_kj ¹²(t) + prod_kj ²³ (t) + prod_kj ¹³(t)] , a3 and b3 being constants.

8. Method according to Claim 7, wherein a3 = 1/2 and b3 = 1/3.

9. Method according to Claim 4, wherein N_SENS= 4 and, for each reference point j (10) , a resemblance function V_kj (t) is calculated, chosen from:

- V_kj(t) = a4.prod_kj ¹²³⁴Ct) ,

- V_kj(t) = b4. [prod_kj¹² (t) + prod_kj ³⁴ (t) ] ,

- V_kj(t) = c4. [prod_kj¹² (t) + prod_kj²³ (t) + prod_kj ³⁴(t) + prodkj¹⁴ (t)] , a4, b4 and c4 being constants.

10. Method according to Claim 9, wherein a4 = 1, b4 = 1/2 and c4 = 1/4.

11. Method according to any one of Claims 4 to 10, wherein the validation parameter is chosen from:

- MAXIMUMo_j = |V_kj (t=0) I ,

- MAXIMUM₁J = Max (|V_kj (t) | ) ,

MaX(MAXIMUM₀.)

- CONTRAST_1J=- ; — —

⁽EjMAXIMUM_1J-MaX⁽MAXIMUM_1J ⁾⁾Z⁽NREF-I⁾ - ENERGY = E₁ (∑_t [s_ki(t)]²) .

12. Method according to Claim 11, wherein only V_kj (t) is calculated for the time t = 0 roughly corresponding to the impact, as being the real part of PROD_kj ⁱ ₁ ⁱ ₂'^Λ2_P(ω) , and MAXIMUMoj is used as validation parameter.

13. Method according to Claim 11 or Claim 12, wherein at least one validation criterion is used, chosen from the following criteria:

- CONTRAST₁J > THRESHOLD₁, with THRESHOLD₁ > 1,

- CONTRAST₁J > THRESHOLD₁ and

MAXIMUM₀J / MAXIMUM:i.j^N0ISE > THRESHOLD₂, where THRESHOLD₂ > 1 and MAXIMUM_1:j^N0ISE corresponds to the parameter MAXIMUM₁J of the signals processed previously and not having resulted in validation, - MAXIMUMoj > THRESHOLD₃, with THRESHOLD₃ > 0 and

MAXIMUMoj / MAXIMUMoj^N0ISE > THRESHOLD₄ with THRESHOLD₄ > 1, where MAXIMUM₀j^NOISE corresponds to the parameter MAXIMUMoj of the signals processed previously and not having resulted in validation,

- MAXIMUMoj / Average (MAXIMUMoj^N0ISE) > THRESHOLD₅, with THRESHOLD₅ > 1,

- ENERGY / ENERGY^N0ISE > THRESHOLD₆, where THRESHOLD₆ > 1 and ENERGY^N0ISE corresponds to the parameter ENERGY of the signals processed previously and not having resulted in validation.

14. Method according to any one of the preceding claims, wherein the reference signals are predetermined theoretically.

15. Method according to Claim 14, wherein the reference signals are used with said object (5) without training phase.

16. Method according to any one of Claims 1 to 13, wherein the reference signals are previously learned on a reference device identical to said object (5) , then are used with said object (5) without training phase.

17. Device specially adapted to implement a method according to any one of the preceding claims, this device comprising: - an object (5) provided with at least N_SENS acoustic sensors (6) , N_SENS being a natural integer at least equal to 2, to pick up respectively N_SENS signals s_ki(t) from acoustic waves generated in the object (5) by an impact on a surface (9) belonging to said object, i being an index between 1 and N_SENS which denotes the corresponding sensor (6) ,

- means for calculating, for each reference point (10) of index j , at least one validation parameter representative of at least one intercorrelation of complex phases representative of said picked up signals S_ki(t) and of reference signals r-ji(t), each reference signal r-ji(t) corresponding to the signal that would be received by the sensor i (6) in case of impact at a reference point j (10) out of N_REF reference points (10) belonging to said surface

(9) , N_REF being a natural integer at least equal to 1 and j being an index between 1 and N_REF,

- and means for locating the impact by determining at least one reference point (10) as close as possible to the point of impact, by applying at least one validation criterion to the validation parameter, characterized in that said validation parameter is representative of at least one intercorrelation: PRODKjV₂'^-12p(ω)=φSui(ω)φRjn(ω)^* φSki₂(ω)^* φR_ji2(ω) ...φS^^(ω)^* φRji2p(ω) where: - φSκi(ω) is the complex phase of S_ki(ω), for i =

- øR_ji(ω) is the complex phase of Rji(ω) , for i =

- ^* denotes the conjugate operator, applied to φS_ki(ω) for i = i_2m and to φRji(ω) for i = i_2m-i, m being an integer between 1 and p;

- S_ki(ω) is the Fourier transform of S_ki(t),

- R_ji(ω) is the Fourier transform of r-ji(t),

- p is a non-zero natural integer less than or equal to N_SENS/2,

- i-_L, i₂, ... i₂p are 2p indices denoting 2p sensors (6) , each between 1 and N_SENS-