FR3144682A1 - Electric switch with gesture control and without contacts. - Google Patents

Abstract

A button, a switch, a slide, or even a lever... So many basic physical control interfaces on which an individual physically acts and which make it possible to open, close and choose between several active states of an electrical device. They are at the origin of more elaborate standard HMIs such as keyboards or mice which only perform a single mechanical action: for example, to turn on a light bulb, an individual presses a switch. They have become essential and today equip many homes and public places. They are also found in most digital digicodes, distributors and even self-service machines. Despite their repeatedly demonstrated effectiveness, their use in public places calls into question a concern for hygiene but also for safety for those who are connected to a larger electrical current. The invention described below relates to a new type of contactor consisting of an optical device and a programmable electrical circuit. This assembly is developed according to a wired and connected electrical architecture. The established device analyzes a scene, and searches for a particular skin-colored image, then activates or deactivates, without contact, associated electrical control processes such as switches, shareholders or electrical switches.

Description

Electric switch with gesture control and no contact.

It is a new device for direct interaction between a human being and an electrical or electronic interconnection network on which an electric current flows. It is associated with spatial and temporal information and can be used to replace some conventional manual control devices for future applications.

Today, to interact with all the electrical devices in our daily environment, there are three approaches:

On the one hand, the classic approach, which consists of acting by pressure on a surface to move a contact device from position A to position B, thus cutting or letting the current (or any flow) pass. In some cases the contact action is direct, in others, the action is performed via a remote control or a module that exchanges information with an electrical system, which allow remote interaction with all the devices in our daily environment.

This approach, which has already proven to be effective, has two major drawbacks: the action leading to intimate contact of the finger with the surface, increases the risk of electrocution if the electrical switches are in a ground fault condition, and also represents a significant potential danger of contamination in the event of an epidemic.

On the other hand, the approach of gestural interfaces, or human-machine interaction based on artificial vision. These electronic devices are equipped with one or more optical sensors that interact with a computer or powerful calculators. One or more image processing algorithms are implemented to detect the hand, segment it, recognize a sign, then associate it with an action.

For better results, it is possible to use instrumented gloves, equipped with numerous sensors measuring the angles of the joints and the position of the hand in space. Particularly used to control virtual actors, by designating and manipulating graphic objects on computer screens [Baudel93, Bellik96] they can also be used for sign language recognition [Braffort96]. Unfortunately, such gloves remain at the very least inaccessible because they are expensive and fragile, and their connecting cables are a hindrance.

Gesture recognition systems have been attempted, for example US patent 5,534,917. Other more recent techniques aim instead at implementing very robust algorithms, which can use Kalman filters with Gaussian assumptions, associated with the common image subtraction technique with the background image. This reference image is then updated automatically, the aim being to obtain a consistent and real-time image processing. However, these systems generally involve complicated and intensive computer analysis and require expensive and bulky hardware.

Other systems use color properties for image analysis. The scene is segmented according to the skin color model. This gesture recognition system has special requirements for accurate capture of the shape and contour of the hand. This implies a uniform background, fixed lighting, high resolution (or multiple cameras) and considerable processing power. In addition, no parasitic movements other than that of the hand must be detected in the scene. These constraints will allow the precise determination of the shape and location of the hand by matching with already recorded models.

The main obstacle of this approach is the separation of the acquisition and processing device. The information is processed pixel by pixel, resulting in a large amount of data to be processed. Segmentation requires very expensive mathematical calculations in terms of computing time. In addition, color classifications, segmentation and optimization are carried out in three different programs (algorithms).

Classical classifiers used by these systems, such as Hidden Markov Models and Neural Networks require heavy preprocessing. Therefore, these systems cannot operate with low-cost, low-power equipment. They are relatively expensive and require a large number of gesture samples to converge. The available interfaces use cameras (CMOS, CDD) and a computer, bulky elements that do not lend themselves to miniaturization.

Finally, we find the presence detection system. It is commonly used in sanitary elements, such as washbasins, faucets, urinals, showers, hand dryers, soap dispensers, etc. It is a control device that allows these devices to be triggered without requiring manual contact from the user. Such devices have the advantage of keeping the sanitary element clean and not touched by the user, which limits the development and transmission of germs or bacteria. In addition, they also make it possible to limit the amount of water or electricity consumed by the sanitary element, since the latter only works in the presence of a user, and stops when the user moves away and no longer uses it.

It is therefore common to equip these sanitary elements with contactless devices, using infrared transmitters and sensors. The control device triggers the operation of the sanitary element according to the infrared intensity detected. If the intensity increases, this corresponds to the presence of a user near the detector. The example of such a device is described in particular in application FR 2 761 446 A1. There are, among other things, infrared detectors that can detect the presence of a user without necessarily being mounted opposite a specific surface. These detectors are background suppression sensors, which, thanks to a system of lenses and a matrix of detectors, allow the distance separating them from the user to be estimated more precisely than with standard infrared sensors. The background suppression sensors can therefore be mounted elsewhere than at the free end of the tap, such as at its base.

All of these devices operate automatically, resulting in a high rate of unwanted, untimely triggering.

It is clear that there is no solution that can equip conventional switches and give them the ability to interact remotely with a human. Existing interaction devices are too heavy, power-hungry, and bulky to replace switches. Presence detectors operate automatically and do not interact with humans. The solution involves algorithmic and architectural adequacy.

A switch as symbolized in figure (1) below, is a device having at least two conductor cable paths (wire, bus) with terminals at the ends that are used to establish electrical contact. Between these terminals? there is an activation device (or switch). This mechanism is covered with a surface on which an individual exerts manual physical pressure that establishes (or breaks) the connection. This operation, generally carried out by the finger of the hand, authorizes or interrupts the circulation of the electric flow between the terminals.

To provide an answer to the problem of hygiene and safety

in public places equipped with this type of installation, the present invention proposes to replace the manually switched electrical (electronic) circuit with a remotely controlled electrical device. The new device would replace the manual switch ( ) by an electrically activated switch (EAS).

The mechanical action of the finger is replaced as in figure (3) by a device (CAE), which then receives an electrical command and automatically modifies the configuration of the two terminals of the switch. The CAE device of the present invention is a basic electric cell or an association of basic cells which receives a voltage on one side and allows the current to pass on the other side. This device (see ) is in turn controlled by a second device, an image processing module (MTI: Image Processing Module).

The present invention proposes to use an image processing module to provide the voltage control of the CAE cell. This module (MTI) has a dual functionality: detect an object in the shape of a finger and that it is the color of a human's skin. Then, it must be able to implement skin color segmentation algorithms and at a useful rate. It is like the ALUs of DSPs, these processors are based on a generalist architecture ( ) capable of performing the operations underlying the information processing primitives. Finally, managed by a sequencer, and at the cost of a few memorization elements, such an approach makes it possible to chain instructions into programs.

An image acquisition tool (see ) completes the entire device of the present invention. It is a network of standard resolution image sensors that can consist of a simple CCD (Charge Coupled Device), CMOS APS (Active Pixel Sensor) or other type camera. This device, interacting remotely with the environment, makes the physical and mechanical action visual. The images filmed continuously by this system are transferred to the image processing module which subsequently generates an electrical command that will activate (or not) the CAE switch.

The final device of the present invention proposes two architectures to address the problem of hygiene, safety, power consumption, computing power, and size: The first architecture is composed of an acquisition and processing module connected to a network of N CAEs and each itself associated with an auxiliary circuit, as illustrated in figure (7). One of these architectures uses a compact circuit of the VLSI (Very Large Signal Itegration) type, consisting of an image sensor matrix, an image processing device, a command-control interface, and possibly a radio frequency circuit depending on the interconnection mode. This compact circuit simultaneously performs image acquisition and processing on the one hand, as well as the control and command of the CAEs on the other hand. It will be produced in standard CMOS technology of low cost and low energy consumption. The processing device used is a DSP (Digital Signal Processing), used according to the operating mode of the present invention is capable of translating the skin color segmentation algorithms and tracking the movement of the finger in a scene. Once the image is processed, the MTI module decodes the information and transfers it via a wired connection, Wifi, or a local radio loop. A voltage command is provided to the corresponding CAE to activate or deactivate one of the ancillary circuits.

The second architecture is a balance on the one hand between electrical consumption, computing power, and hygiene and safety on the other. This architecture is intended to replace the electrical wiring and diagram in the home. Thus, the present invention proposes a single calculator to process images from several cameras, and to have the computing power necessary to process several image sources at the same time and at a useful rate. Then to have cameras in place of switches so as to have a network of sensors whose only functions are to acquire and transmit images. (See ). For this, the sensors are connected to the image processing module by a wired connection, or via a wireless transmission device.

The MTI is in turn connected to the CAEs, as before, by a wired or wireless connection. For each image received and processed by the MTI, an activation code for one (or more) CAEs is generated via the connections which will then activate the associated auxiliary circuits. Each CAE is interconnected to a control circuit and possibly equipped with a radio frequency circuit. This architectural organization makes it possible to increase the computing power of the MTI, to relocate the processing and to reduce the cost of copper if necessary.

Feasibility of an MTI.

One of the preferred algorithms of this invention is a combination of two approaches: The YCbCr chrominance space and the subtraction of the image from the background by temporal modeling of each pixel with a Gaussian mixture and this in an unsupervised manner. For segmentation, the algorithm used is derived from K-means. To simplify, the algorithm is described for a black and white image, it can easily be adapted for an RGB image: by replacing the standard deviation by a covariance matrix, and the intensity of pixel I by a vector of (R,G,B)……

1. Etape 1

%Donner une valeur à la constance temporelle 1/α du processus %d’oubli, donner une valeur faible au poids Ω et grand à l’écart %type σ pour forcer la convergence, T paramètre bimodale ces %valeurs étant issue de %l’apprentissage.
Algorithm

1. Step 1

%Give a value to the temporal constancy 1/α of the forgetting process, give a low value to the weight Ω and a large value to the standard deviation σ to force convergence, T bimodal parameter these %values coming from %learning.

α = α0; T=T0; Ω=Ω0 ; σ=σ0 ;

2) Step 2

%Initialize the parameters of the three Gaussians of the mixture of %each pixel. One Gaussian for the background image, one in case of %bimodal background image, one for the background change over time.

For each Gaussian of each pixel do

For k=1 :3

Ω(k,t=0)=0 Ω(k,t=0)=0 ;

σ(k,t=0)=σ0 𝜎(k,t=0)=𝜎0 ;

µ(k,t=0)=0 µ(k,t=0)=0 ;

end for

%σ is the variance, Ω is the weight, µ is the mean, t is the %time.

End For each Gaussian of each pixel do

3) Step 3

% Calculation of the Gaussian parameters for the current image at %time t

For all pixels of intensity I(t) at time t

For each pixel do

For k=1 : 3

Find k which minimizes this function;

abs(µ(k,t)−I(t))/σ(k,t)abs(µk,t−It)/𝜎(k,t);

k=k0;

End for

if(abs(µ(k0,t−I(t))≤2.5∗σ(k0,t)absµk0,t−It≤2.5∗𝜎(k0,t))

%Update parameters associated with this Gaussian.

µ(k0,t)=(1−α)µ(k0,t−1)+αI(t); µk0,t=1−𝛼µk0,t−1+𝛼I(t);

σ2(k0,t)=(1−α)∗σ2(k0,t−1)+α∗(I(t)−µ(k0,t))2;

𝜎2k(0,t)=1−𝛼∗𝜎2k0,t−1+𝛼∗It−µk0,t2;

Ω(k0,t)=(1−α)Ω(k0,t−1)+α Ωk0,t=1−𝛼Ωk0,t−1+𝛼 ;

If (k≠k0)

Ω(k≠k0,t)=(1−α)Ω(k0,t−1);Ωk≠k0,t=1−𝛼Ωk0,t−1;

End if

ΩN=Ω(1,t)+Ω(2,t)+Ω(3,t) ;

For i=1:3 %normalization of Ω

Ω(i,t)=Ω(i,t)/ΩN;

End for

% the reference image of the pixel

Iref=µ(k0,t) ;

Else

For i =1:3

Calculate the largest ratio Ω(i,t)/σ(i,t)

i0=i ;

End for

% initialize a new gaussian

µ(i0,t)=I(t)µ(i0,t)=I(t);

σ(i0,t)=σ0σ(i0,t)=𝜎0;

Ω(i0,t)=Ω0;Ω(i0,t)=Ω0;

%the future reference image

Iref=µ(i0,t) ;

End else

End if

Image = I(t)- Iref(t-1)

% image subtraction runs with background image

If (Cr0<Cr<Cr1 and Cb0<Cb<Cb1)

%YCrCb chrominance space

So Skin color image;

End

End For all pixels of intensity I(t) at time t,

4) Step 4

%Detection of the direction and the sense of variation of the finger on the scene.

Detected if finger is present on scene;

Detects the direction and direction of finger movement: ;

case 0: no movement

case 1: from right to left

case 2: from left to right

case 3: from bottom to top

case 4: from top to bottom

case 5: diagonal from bottom left to top right

case 6: diagonal from high right to low left

case 7: diagonal from top left to bottom right

case 8: diagonal from bottom right to top left

Analysis of the algorithm.

• Une phase d’étiquetage,
• Une phase de mise à jour des paramètres des clusters (µ, σ et Ω),
• Une phase de normalisation.

A l’issue de cette phase, les caractéristiques géométriques des 3 clusters sont extraits (barycentre, surface, longueur, largeur).
The labeling algorithm is iterative, each iteration is broken down into 3 phases:

• A labeling phase,
• A phase of updating the cluster parameters (µ, σ and Ω),
• A phase of normalization.

At the end of this phase, the geometric characteristics of the 3 clusters are extracted (barycenter, surface, length, width).

Iterative phase of segmentation:

Pixels are labeled into k=3 categories. The category assignment of each pixel is done based on the distance to the target value (µ) of the category.

Once the pixels are labeled, the parameters µ, σ and Ω of each of the k=3 categories are updated according to a recursive average if the distance of the target value to the gray level (or chrominance) value is less than an amount proportional to the σ of the cluster according to the following principle:
IF (
) :

ENDIF

Post segmentation phase:

The geometric characteristics (length, width, barycenter, surface area, etc.) of each of the clusters are calculated.

From this data, the finger position calculation is updated. The number of elementary operations required is approximately 2-3 times the number of pixels.

Assessment of the number and types of operations required

Assumptions: The target architecture is given below:

The operations are performed at the foot of the column by a dedicated SIMD (Single Instruction Multiple Data) processor vector. The elementary processor instruction set is optimized to perform the calculations.

We will assume that the divisions can be performed in 1 operation.

We will assume that the convergence of the algorithm is reached after 8 iterations.

We will assume that the imager is monochrome and uses one of the formats below:

[tab 1] RxC Imager R C CGA 200 320 QVGA 240 320 VGA 480 640 Table 1 Standard formats of low-cost imagers

A monochrome imager can pick up a format like this.

We will assume operation at a rate of 50 or 60 frames/s to avoid stuttering phenomena with artificial lighting.

Results: We then arrive at the table below.

[Tab 2]

type of operation number of pixels number of clusters number of ops degree of parallelization possible Labeling subtraction/pixel/cluster RxC k RxCxk C division/pixel/cluster RxC k RxCxk C comparison/pixel/cluster RxC k RxCxk C assignment/pixel/cluster RxC k RxCxk C storage/pixel/cluster RxC k RxCxk Update comparison/pixel/cluster RxC k RxCxk 2 multiplications x 3 parameters (µ, sigma, Omega) RxCx2x3 k RxCxk 1 addition 1 division/cluster Others Length/width
conditional increment/pixel/cluster RxC k RxCxk C Surface conditional increment/pixel/cluster RxC k RxCxk C Barycenter Multiplication/accumulation RxC k RxCxk C Table 2: Census of the number of operations

Running at 50 or 60 frames/s we could achieve results like these.

For a CGA format and a frame rate of 50fps, we arrive at a total of 500 MOps, or a total of 2M instructions/s with the target architecture.

For a VGA format and a frame rate of 50fps, we arrive at a total of 2 GBps, a total of 4M instructions/s with the target architecture.

Using an IcyCam® type solution

Architecture reminder The architecture of the IcyCam is given below

• 32-bit DSP / MCU icyflex1 processor
• 64-bit data bus
• Contrast and Orientation computation in the data readout path
• Graphical Processing Unit (GPU)
• 32-bit 100 MHz SDRAM interface
• 12-bit Parallel Peripheral Interface (PPI)
• 2 Serial Peripheral Interfaces (SPI)
• 16-bit General Purpose I / O Interface (GPIO)
• UART interface
• JTAG interface
• Programmable in C (GNU tool suite)
• Max. system clock rate : 33 MHz

The main features are:

• 32-bit DSP/MCU icyflex1 processor
• 64-bit data bus
• Contrast and Orientation computation in the data readout path
• Graphical Processing Unit (GPU)
• 32-bit 100MHz SDRAM interface
• 12-bit Parallel Peripheral Interface (PPI)
• 2 Serial Peripheral Interfaces (SPI)
• 16-bit General Purpose I/O Interface (GPIO)
• UART interface
• JTAG interface
• Programmable in C (GNU tool suite)
• Max. system clock rate: 33 MHz

Analysis of processing capacities

Considering the architecture, we can estimate the maximum computing capacity at 50MIPS. The IcyCam allows processing at a safe video rate for simple algorithms (optimal thresholding, contour detection by linear filtering and thresholding, etc.). On the other hand, for a k-means segmentation algorithm, this architecture is limited.

To overcome this limitation, the target architecture must allow a decimation of the total number of pixels and an increase in the degree of parallelism of the photosensitive matrix.

The image processing circuit of the present invention uses the principle of a target architecture. An architecture that brings together in a single ASIC circuit a photo sensor matrix, a column processor vector, storage elements, and a microcontroller. Then, to increase the degree of parallelism, the target architecture integrates around each sensor, the skin color image segmentation function in the chrominance space (R, G, B). This equation is described below:

(R > 95) AND (G > 40) AND (B > 20) AND (MAX(R, G, B) - MIN(R, G, B) > 15) AND (ABS(R-G) > 15) AND (R > G) AND (R > B). This amounts after simplification to: (R > 95)&(G > 40)&(B > 20) &(R > G+15)&(R>B).

Then the algorithm must manage the decrease in the decimation of the pixel matrix: 1) by reducing the number of columns by any factor X, 2) by dividing the number of images processed per second by a factor Y. That is a total decimation by a factor X.Y of the total number of pixels.

Description of the embodiments

An example of realization according to the model of architecture1: The developed electrical diagram of a double ignition installation in a home. The two switches are replaced by two finger movements to be performed in front of a camera. The gestures are to be performed between 10 cm and 2.5 m from the front of the camera and following an angle (+-) 60 degrees from the perpendicular to the lens. One movement to turn on and another to turn off, i.e. four movements in total as illustrated in figure 20 to achieve the double ignition.

Example of Implementation of a CAE: One of the methods of implementing the CAE is illustrated in the diagram of the , 14 this is an electrical relay controlled by a voltage. Activation is done by providing a voltage peak Vd on terminals c and d, the switch goes from L1 to N or vice versa.

Example of the Creation of an MTI

A possible embodiment of an MTI of the present invention from commercial electronic elements, uses a standard computer equipped with a camera, a USB port, a parallel port, a serial port. By “standard”, we mean a PC where the editing software, communication software, as well as the various peripheral drivers are installed. In this PC, the image processing software is installed, on its USB port is connected a 5Volt/230Volt electrical relay, and on the latter is interconnected the phase and the “phase return”, on one side the phase comes from the electrical network of a home and the “phase return” is connected to a 220Volt lamp as well as the associated neutral. And using the Matlab software, the target image processing algorithm was implemented. This is a program that detects the variation of a hand finger in four orientations: from bottom to top, from top to bottom, diagonally from bottom to top, and diagonally from top to bottom. This corresponds to four control actions, respectively, turning on, turning off, increasing, decreasing the light. According to another embodiment, the variations of the above movement can be reversed and a total of 8 actions would be obtained.

And if we add the finger pointing gesture to these 8 variations, we get an 8 to 9 digit “digicode”.

ILLUSTRATIVE DIAGRAMS OBTAINED BY MATLAB SIMULATION

There illustrates the part of the algorithm dealing with the case of the hand in an ascending phase. The hand would have to be in an ascending phase for at least 10 successive images and would have to reach the top of the screen for the increase in the intensity of the lamp to be triggered.

There illustrates the part of the algorithm dealing with the case of the hand in a descending phase. The hand would have to be in a descending phase for at least 10 successive images and would have to reach the bottom of the screen for the decrease in the intensity of the lamp to be triggered.

There illustrates the part of the algorithm dealing with the case of the hand in a diagonal upward movement phase. The hand would have to be in a diagonal upward movement phase for at least 10 successive frames and would have to arrive towards the extreme left of the top of the screen for the lamp to be triggered.

There illustrates the part of the algorithm dealing with the case of the hand in a downward diagonal movement phase. The hand would have to be in a downward diagonal movement phase for at least 10 successive frames and would have to arrive towards the far right of the bottom of the screen for the lamp to be turned off.

:

This figure represents the symbols of switches open, another closed. The two vertical half-lines symbolize the action of opening and closing. And the middle part the switch or presses to turn on or off a lamp.

:

This figure shows a common manual switch with its switch in the center allowing you to switch from one mode to another.

:

This figure symbolizes a system capable of acting remotely to open or close a switch. The device receives a voltage and acts in place of human mechanical action to interconnect the 2 electrical wires.

:

This figure shows an auto electric switch (AES) connected to the Image Processing Module (IPM) which provides a VD control signal.

:

This figure represents the general architecture of the image processing device with the essential elements. It includes processors, a pixel plane, a microcontroller, a communication bus interface, and the mixed processor lines.

:

This figure shows the CAE interconnected to the MTI by a wired connection. The MTI is connected to the image acquisition device. The latter sends the images to the MTI which processes them and provides the command and control of the CAE, which will activate or not the electrical flow.

:

This figure represents the overall device of the invention according to the embodiment of architecture 1. It shows the assembly between the camera with the image processing module and the auto electric switch in a single electric circuit.

A compact printed circuit board or, the image acquisition device and the MTI will be realized in low cost Very Large Signal Integrated (VLSI) CMOS ASIC.

:

This is the target architecture of the image processing device of the invention. This architecture is derived from the target algorithm of the invention. It is a matrix of photo sensors, processor columns, temporary storage elements and global memory planes.

:

Reminder of the standard Icycam type image processing architecture. It is an architectural solution that uses memory RAMs and a 320*240 QVGA camera associated with an Icyflex type processor.

:

This figure represents an Icyflex type image processing processor architecture with memory interfaces and logical calculation units.

:

This figure shows all the devices of the invention in an electrical diagram of the dwelling and according to the embodiment of architecture 2.

:

This figure represents the use of the device of the invention within the electrical installation of a home for a double ignition. An individual can communicate with the device from a distance of 20 cm to 2 m.

:

This figure represents an example of CAE. It is an open or closed electrical relay controlled by voltage VD.

:

This figure shows a lamp connected to a relay with a voltage of 5 volts. This is an example of a CAE implementation with a 5/230V relay.

:

This figure represents the result of the target algorithm implemented in a PC that tracks the movement of the finger of the hand from bottom to top.

:

This figure represents the result of the target algorithm implemented in a PC that tracks the finger movement from top to bottom.

:

This figure represents the result of the target algorithm implemented in a PC that tracks the movement of the finger from the low right diagonal to the high left.

:

This figure represents the result of the target algorithm implemented in a PC that tracks the movement of the finger from the top right diagonal to the bottom left.

: The four directions of movement to turn two lamps on and off

Claims (8)

Switch device which according to the invention uses a remote-controlled electrical device to authorize or interrupt the flow of electrical flow between two conductive cable paths (wire, bus) comprising two ends, terminals which are used to establish electrical contact. Said device consists of:

• From an electrically activated switch (EAS) device whose activation is done remotely,
• Of an image processing module (IPM) device whose function is to segment the movement of the finger of the hand and to give the order to the electrically activated switch to authorize or interrupt the electrical flow,
• From an image acquisition tool device that films the scene continuously,
• Of a network device of interconnections which are wired or wireless links.

The switch device of claim 1, wherein said electrically activated switch receives an electrical command and automatically changes the configuration of the two terminals of the switch,
Characterized in that said device (CAE) of the present invention is a basic electric cell or an association of basic cells which receives a voltage on one side and lets the current pass on the other. The switch device according to claim 1 and 2, wherein said activation command of said electrically activated switch is provided by the image processing module device. Said image processing module device uses an object segmentation algorithm of the shape of a finger. Then said device detects the variation of the finger according to the predefined orientations and activates said associated commands. The switching device according to claim 1, wherein
The image acquisition tool device is a network of image sensors of any resolution that can consist of a simple CCD (Charge Coupled Device), CMOS APS (Active Pixel Sensor) or other type camera. Said tool films the scene continuously or on demand. Said filmed images are transferred to the image processing module device which, after processing, decodes the control signals of the electrically activated switching devices. The switch device of the invention according to claim 1, in which the acquisition tool device, the image processing module device, and the command/control interface device have been produced in the same compact circuit of the VLSI (Very Large Signal Itegration) type. Said compact circuit simultaneously performs the acquisition and image processing on the one hand, as well as the control and command of said electrically activated switches. Switching device of the invention according to claim 5, in which said compact circuit is constituted:
From an image sensor matrix,
Processors,
From a sequencer/microcontroller,
From a control command part,
Characterized by the fact that the processor device is of the DSP (Digital Signal Processing) type, capable of translating skin color segmentation algorithms and tracking the movement of the finger in a scene. Switch device according to claim 1, in which the device networks interconnections between:
The image acquisition tool device and the image processing module device are connected by a wired connection, WiFi, or a local radio loop.
The image processing module device and the electrically activated switch device are also connected by a wired connection, Wifi, or a local radio loop.
The electrically activated switching device is directly connected on either side of the same wire of said network. Switch device of the invention according to claim 1, in which, The image processing module device is a power calculator connected to several acquisition tool devices. Said calculator has the necessary computing power to process all the images at a useful rate: Characterized by the fact that image acquisition tools can be provided in place of and in place of the switches, so as to have a network of sensors whose only functions are to acquire and transmit the images to the image processing module devices.