Loudspeaker comprising electronic detection device for detecting wall
By integrating loudspeakers, microphones and electronic detection devices in the speakers, using neural network models to detect wall positions and adjust broadcast modes, the problem of sound delay when the speaker approaches the wall is solved, and the user experience is improved.
Patent Information
- Application Number
- CN202411366649.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Priority Date
- 2023-10-03
- Filing Date
- 2024-09-29
- Publication Date
- 2025-04-04
AI Technical Summary
When the speaker is placed close to the wall, the emitted sound transmission effect is affected by obstacles, resulting in sound delay and user experience degradation. The signal transmission needs to be adjusted to limit the decline of user experience.
A speaker including a loudspeaker, multiple microphones and electronic detection devices is used to calculate the spectrogram and energy level of the component arrival direction of the audio signal through the calculation module, and a neural network model is used to determine the presence or location of the wall in the speaker environment to adjust the broadcast mode.
It realizes accurate detection of wall positions in speaker environment and automatic adjustment of broadcast mode, improving user experience and reducing sound delay.
Smart Images
Figure CN119767219A_ABST
Abstract
Description
[0001] The present application relates to a loudspeaker comprising at least one loudspeaker capable of emitting a transmitted audio signal, and a plurality of microphones, each microphone being capable of acquiring an audio signal received by the loudspeaker.
[0002] The present application relates to the field of loudspeakers, preferably portable loudspeakers (also called mobile loudspeakers).
[0003] Such loudspeakers can usually be placed in different environments presenting different structures.
[0004] The acoustic performance of the sound emitted by a loudspeaker is strongly affected by the environment, in particular by the presence of obstacles such as walls. In fact, when a loudspeaker is placed close to a wall, for example less than 50 cm from the wall, the emitted sound is reflected from the wall and propagates to the user with a delay relative to the sound along the direct path between the loudspeaker and the user.
[0005] This results in a degraded experience for the speaker users.
[0006] Therefore, there is a need to determine the location of walls in the loudspeaker environment in order to be able to adjust the signal emission from the loudspeaker in order to limit degradation of the user experience.
[0007] To this end, the present application is directed to a loudspeaker further comprising at least one loudspeaker capable of emitting an emitted audio signal,
[0008] - a plurality of microphones, each microphone being capable of acquiring a received audio signal, if there are walls in the speaker environment, the received audio signal comprising the transmitted audio signal that has been reflected by the wall, and
[0009] - electronic detection equipment for detecting the wall connected to each microphone,
[0010] Electronic testing equipment includes:
[0011] a first calculation module capable of calculating a spectrogram of the directions of arrival of components of the audio signal received by the loudspeaker from the received audio signal acquired by each microphone,
[0012] Characterized in that the electronic detection equipment also comprises:
[0013] a second calculation module capable of calculating the energy level of the received audio signal obtained from each microphone, and
[0014] A determination module capable of determining the presence or absence of a wall in the speaker environment by applying a neural network model to the spectrogram of the directions of arrival and the determined energy levels.
[0015] According to a particular embodiment, the loudspeaker according to the present application comprises one or more of the following features, either alone or according to any technically possible combination:
[0016] - the neural network model comprises a first convolutional neural network capable of receiving a spectrogram of a direction of arrival, a second neural network capable of receiving an energy level, a cascade block capable of concatenating data from the first and second neural networks to form cascaded data, and a third neural network capable of processing the cascaded data to determine the presence or absence of a wall,
[0017] - the first neural network comprises a circular sliding window and at least one neural layer, the sliding window being able to be applied to the spectrum of the directions of arrival before the neural layer,
[0018] - if there is a wall in the environment of the loudspeaker (10), the neural network model is able to determine the position of the wall relative to the loudspeaker from a number of predefined positions based on the spectrum of the directions of arrival and the calculated energy levels,
[0019] - at least one loudspeaker is capable of emitting a transmitted audio signal according to a broadcast mode of the loudspeaker,
[0020] The electronic detection device also includes an adaptation module, which can reconfigure the broadcast mode of the speaker according to the presence or absence of the wall.
[0021] - the spectrogram of the direction of arrival is a matrix containing a plurality of values, each value corresponding to the power of the audio signal received by the loudspeaker according to a predefined angular interval relative to a predefined reference system centered on the center of the loudspeaker, and according to a predefined frequency interval,
[0022] - the transmitted audio signal is included in the audio stream of the broadcast instruction from the user; and
[0023] The speaker also includes an accelerometer capable of determining whether the speaker is stationary and transmitting a calculation instruction when the speaker is stationary, and a first calculation module and a second calculation module capable of calculating the spectrum diagram and energy level of the arrival direction after receiving the calculation instruction from the accelerometer.
[0024] The present application also aims at a method for detecting a wall in a loudspeaker environment comprising at least one loudspeaker, several microphones and an electronic detection device for detecting the wall connected to each microphone,
[0025] The method comprises the following steps:
[0026] transmitting an audio signal by at least one speaker,
[0027] - obtaining the audio signal received by each microphone, if there is a wall in the environment of the loudspeaker (10), the received audio signal includes the audio signal that has been reflected from the wall,
[0028] - calculating, from the received audio signal acquired by each microphone, a spectrogram of the directions of arrival of the components of the audio signal received by the loudspeaker,
[0029] It is characterized in that the method further comprises the following steps:
[0030] - calculate the energy level of the received audio signal acquired from each microphone, and
[0031] -Determine the presence or absence of walls in the speaker environment by applying a neural network model to the spectrogram of the directions of arrival and the determined energy levels.
[0032] The present application also aims at a computer program product comprising software instructions which, when executed by a computer, implement the detection method as described above.
[0033] The present application will become more apparent from a reading of the following description, which is given by way of non-limiting example only and is described with reference to the accompanying drawings, in which:
[0034] Figure 1 is a schematic diagram of a loudspeaker according to the present application;
[0035] Figure 2 is able to be included in Figure 1 An example of a spectrogram of directions of arrival determined by electronic detection equipment in a loudspeaker;
[0036] Figure 3 is able to be included in Figure 1 An example of an electronic detection device determining energy levels in a loudspeaker;
[0037] Figure 4 Is Figure 1 A schematic diagram of a neural network model implemented by an electronic detection device included in; and
[0038] Figure 5 Is Figure 1 Flow chart of the detection method implemented by the speaker in.
[0039] exist Figure 1 In FIG. 1 , a loudspeaker 10 is shown in an environment. The environment comprises a wall 15 near the loudspeaker 10. “Nearby” means a distance less than or equal to 1 m, preferably less than or equal to 50 cm.
[0040] The loudspeaker 10 comprises at least one loudspeaker 20 capable of emitting audio signals. Preferably, the loudspeaker 10 comprises several loudspeakers 20, preferably distributed around the periphery of the loudspeaker 10, to emit audio signals in several directions.
[0041] exist Figure 1 In the example of FIG. 1 , the speaker 10 includes two loudspeakers 20 .
[0042] The speaker 10 also includes a plurality of microphones 25, each of which is capable of acquiring an audio signal present in the environment of the speaker 10. Figure 1 In the example of the speaker 10, the speaker 10 includes four microphones distributed on the outer surface of the speaker 10. The microphones 25 are preferably evenly angularly spaced from each other. Therefore, in the example of the speaker 10 including four microphones 25, the microphones 25 are angularly distributed with an angular displacement equal to 90° so as to acquire audio signals in directions as far away from each other as possible.
[0043] The loudspeaker 10 also includes a control device 27, an accelerometer 28, and an electronic detection device 30. The control device 27 is connected to the loudspeaker 20 and optionally to a source of audio content (not shown), such as a digital player, a compact disc player, a vinyl disc player, or a smart phone. The control device 27 is capable of receiving audio content from the source and sending an excitation control to the loudspeaker 20 to play the received audio content. The excitation control then causes the loudspeaker 20 to emit an audio signal, referred to as a transmitted audio signal. Preferably, the control device 27 is capable of sending the excitation control according to a first broadcast mode of the loudspeaker 10, which is selected from several predefined broadcast modes of the loudspeaker 10. For example, the selected broadcast mode is a neutral mode, in which the loudspeaker 20 is controlled in a synchronized manner.
[0044] It can then be understood that the transmitted audio signal is included in the audio stream from the user broadcasting instructions. In other words, the transmitted audio signal is, for example, any type of music or speech, and is not limited to a predefined test signal.
[0045] If a wall 15 is present in the environment of the loudspeaker 10 , the emitted audio signal is at least partially reflected by the wall 15 .
[0046] Each microphone 25 is then able to acquire the audio signal received by the speaker 10 after the loudspeaker 20 has emitted the audio signal.
[0047] The received audio signal contains a direct component, which is the audio signal transmitted according to a direct path between the loudspeaker 20 and the microphone 25 .
[0048] If there is a wall 15 in the environment of the speaker 10, such as Figure 1In the example in , the received audio signal also contains a reflection component, which is the audio signal transmitted and reflected on the wall 15.
[0049] Optionally, the received audio signal also contains noise. In other words, the received audio signal contains:
[0050] - direct component and noise if there is no wall 15 in the environment of the loudspeaker 10; or
[0051] - If there is a wall 15 in the environment of the loudspeaker 10, it is the direct component, the reflected component and the noise.
[0052] The accelerometer 28 is connected to the detection device 30. The accelerometer 28 is able to measure the acceleration of the loudspeaker 10 and is able to send calculation instructions to the detection device 30 when the loudspeaker 10 is at rest again.
[0053] The electronic detection device 30 comprises a first module 35 for calculating the spectrogram of the direction of arrival, a second module 40 for calculating the energy level, a module 45 for determining the presence or absence of a wall 15 and an optional module 50 for adapting the scattering pattern.
[0054] exist Figure 1 In the example of FIG. 5 , the electronic detection device 30 is a computer, including, for example, a memory 55 and a processor 60 associated with the memory 55 .
[0055] exist Figure 1 In the example of the first calculation module 35, the second calculation module 40 and the determination module 45, as well as the optional addition, the adaptation module 50, are each implemented in the form of software or software bricks that can be executed by a processor. The memory 55 of the electronic detection device 25 can then store the first calculation software, the second calculation software and the determination software, as well as the optional addition, the adaptation software. The processor can then execute each of the first calculation software, the second calculation software and the detection software, as well as the optional addition, the adaptation software.
[0056] As an alternative (not shown), the first calculation module 35, the second calculation module 40 and the determination module 45, as well as the optional addition, the adaptation module 50, are all implemented in the form of a programmable logic component (such as an FPGA (field programmable gate array)) or an integrated circuit (such as an ASIC (application integrated circuit)).
[0057] When the electronic detection device 30 is implemented in the form of one or more software programs, in other words, in the form of a computer program (also referred to as a computer program product), it can also be recorded on a computer-readable medium (not shown). A computer-readable medium is, for example, a medium capable of storing electronic instructions and capable of being coupled to a bus of a computer system. For example, the readable medium is an optical disk, a magneto-optical disk, a ROM memory, a RAM memory, any type of non-volatile memory (e.g., a flash memory or NVRAM) or a magnetic card. The computer program containing the software instructions is then stored on the readable medium.
[0058] The control device 27 is also, for example, a computer including a memory (not shown) and a processor (not shown).
[0059] Alternatively, the detection device 30 and the control device 27 form a single computer comprising a single processor 60 and a single memory 55. The memory 55 then also stores the control software.
[0060] For example, after receiving the calculation instruction from the accelerometer 28, the first calculation module 35 can acquire the audio signal received from each microphone 25. The first calculation module 35 can calculate the spectrogram of the arrival direction of the received signal on the speaker 10 based on the audio signal received by the speaker 10 and acquired by each microphone 25. The spectrogram of the arrival direction is calculated based on all components of the received audio signal.
[0061] To this end, the first calculation module 35 can for example apply the MUSIC (Multiple Signal Classification) algorithm known in the prior art and described in detail in the following webpage https: / / fr.wikipedia.org / wiki / MUSIC_(algorithm). In a manner known per se, the MUSIC algorithm is configured to detect sound signals coming from a search direction. When it detects a sound signal coming from a certain direction, it must be a reflected component, since, by construction, the direct component arrives in the same way from all directions.
[0062] The MUSIC algorithm is then used, allowing the identification of the preferred direction of arrival of the received audio signal.
[0063] If a wall 15 is present in the environment of the loudspeaker 10, the identified components substantially correspond to reflected components of the received audio signal.
[0064] The MUSIC algorithm provides a matrix.
[0065] The columns of this matrix represent the directions of arrival around the loudspeaker 10 according to a first sample of the directions of arrival. Thus, in a predefined reference frame centered on the center of the loudspeaker 10, the columns correspond to angular intervals of equal width from one column to another.
[0066] The rows of this matrix correspond to the frequency bins. For example, the matrix contains 257 rows.
[0067] Each value of the matrix is associated with a column and a row and corresponds to the power of a component identified by the MUSIC algorithm in the angular interval of the associated column and the frequency interval of the associated row.
[0068] The first calculation module 35 can also optionally perform frequency subsampling and smoothing in sixth octaves on the frequency axis of the matrix to reduce the number of rows, for example to 25. Alternatively, smoothing can be implemented in twelfth octaves, one-third octaves or per octave.
[0069] Furthermore, the first calculation module 35 can optionally reduce the number of columns by averaging the values over several angular intervals, for example so that the matrix contains only 8 columns. In other words, after averaging, each column corresponds to an angular interval of 45°.
[0070] Furthermore, the first calculation module 35 is preferably capable of normalizing the values of the obtained matrix so that the highest value corresponds to the value 1, and the other values correspond to percentages of the highest value.
[0071] The direction of arrival with the highest value is therefore the most probable direction of arrival of the reflected component at the loudspeaker 10. This direction of arrival or these directions of arrival are therefore the most probable directions in which the wall 15 is positioned relative to the loudspeaker 10.
[0072] The matrix obtained by the first calculation module 35 is then a spectrogram of the directions of arrival of the sound sources around the loudspeaker 10, including in particular the reflected components identified when they exist. An example of such a spectrogram is Figure 2 In grayscale. Figure 2 , each box corresponds to a value in the obtained matrix.
[0073] exist Figure 2 In the figure, the brighter the box, the higher the power of the reflected component identified on the speaker 10 in the corresponding frequency and angle range. Conversely, the darker the box, the lower the power of the reflected component identified in the corresponding frequency and angle range.
[0074] Figure 2 Two regions 65 are shown. The regions 65 correspond to the group of frequency intervals where the power of the identified reflected components is highest.
[0075] exist Figure 2In the example of , these regions correspond to angular intervals between 90° and 135° on the one hand and between 270° and 315° on the other hand. Therefore, there is an ambiguity of 180° regarding the direction of the wall 15 relative to the loudspeaker 10. This ambiguity is due to the fact that the MUSIC algorithm is designed to calculate the spectrogram of the direction of arrival in a point source environment, rather than the spectrogram of the reflection of the audio signal on the surface formed by the wall 15.
[0076] Furthermore, the material of the wall 15 has an influence on the reflected frequencies of the emitted audio signals. In practice, depending on the material of the wall 15, certain frequencies are mainly absorbed, while other frequencies are mainly reflected.
[0077] The second computing module 40 can also receive audio signals from each microphone 25 , for example after receiving computing instructions from the accelerometer 28 .
[0078] The second calculation module 40 is capable of calculating the energy level of the received audio signal acquired from each microphone 25 .
[0079] To this end, the second calculation module 40 can calculate the energy level of the received audio signal acquired by each microphone 25 within a predefined duration (e.g. 5 seconds). Preferably, the calculation module 40 can calculate the energy level only on a predefined frequency band (e.g. between 20 Hz and 20,000 Hz).
[0080] For example, the energy level is calculated using the following formula:
[0081]
[0082] Where: E i is the energy level received by microphone i,
[0083] T is a predefined duration,
[0084] f min and f max are the minimum and maximum frequencies of the band,
[0085] X i (t,f) is the audio signal received by microphone i at time t and frequency f in the frequency domain, and
[0086] conj(.) is the complex conjugate function.
[0087] The second calculation module 40 can normalize the energy levels such that the highest energy level is equal to 1 and the other energy levels are equal to percentages of the highest energy level.
[0088] Figure 3 shows an example of calculating energy levels. Figure 3It can be seen that the first microphone 25 picks up the received audio signal with the highest energy level, ie a normalized level equal to 1. It can also be seen that the other microphones 25 pick up the received audio signal with lower energy levels, ie normalized levels between 0 and 1.
[0089] Reference again Figure 1 , the determination module 45 is connected to the first calculation module 35 and the second calculation module 40.
[0090] The determination module 45 is able to determine the presence of a wall 15 or the absence of a wall 15 in the environment of the loudspeaker 10 by applying the neural network model 70 to the spectrogram of the directions of arrival and the determined energy levels.
[0091] Figure 4 The architecture of the neural network model 70 implemented by the determination module 45 is illustrated.
[0092] The neural network model 70 includes a first convolutional neural network 75 capable of receiving a spectrum diagram of an arrival direction, a second neural network 80 capable of receiving an energy level, a cascade block 83 intended to cascade data from the first neural network 75 and the second neural network 80 to form cascaded data, and a third neural network 85 capable of processing the cascaded data to determine the presence or absence of a wall 15.
[0093] Each neural network consists of a series of ordered neural layers, with the input of each neural layer being obtained from the output of the previous layer.
[0094] More precisely, each layer contains neurons that take input from the outputs of neurons in the previous layer or even from the input variables of the first layer.
[0095] Each neuron is also associated with an operation, in other words, the type of processing performed by the neuron in the corresponding processing layer.
[0096] Each layer is connected to other layers through multiple synapses. A synaptic weight is associated with each synapse, and each synapse forms a connection between two neurons. Each synaptic weight is preferably a real number, which has positive and negative values. In some cases, each synaptic weight is a complex number.
[0097] Each neuron is capable of performing a weighted sum of the values received from the neurons of the previous layer, then multiplying each value by the respective synaptic weight of each synapse and connection between the neuron and the neurons of the previous layer, then applying an activation function (usually a nonlinear function) to the weighted sum, and passing the value resulting from the application of the activation function at the output of the neuron, in particular to the neurons of the next layer connected to it. The activation function allows nonlinearity to be introduced into the processing performed by each neuron. The sigmoid function, the hyperbolic tangent function, the Heaviside function, the rectified linear unit (ReLU) function or the SoftMax function are examples of activation functions.
[0098] As an optional addition, each neuron can apply a multiplicative factor (also called bias) to the output of the activation function, and the value delivered by said neuron on the output is then the product of the bias value and the value from the activation function.
[0099] Convolutional Neural Networks are sometimes also called Convolutional Neural Networks (CNN).
[0100] In a convolutional neural network, each neuron in the same layer has exactly the same connection pattern as its neighboring neurons, but at different input locations. The connection pattern is called a convolution kernel.
[0101] A fully connected neuron layer is a layer in which each neuron of the layer is connected to all neurons of the previous layer.
[0102] This type of layer is often called a "fully connected" layer, or sometimes a "dense layer".
[0103] The first neural network 75 comprises a circular sliding window (circular filling) and at least one neural layer. Preferably, the first neural network 75 comprises three layers of neurons.
[0104] The sliding window contains a kernel of width M and height N, where M is less than the number of columns in the matrix forming the spectrogram and N is less than the number of rows in the matrix. The window can be moved in the spectrogram to form sub-matrices of M by n dimensions. These sub-matrices are provided to at least one neural layer.
[0105] The circular shape of the sliding window enables the window to form a submatrix containing values from the first and last columns. In other words, the sliding window moves over the spectrogram as if the first column of the corresponding matrix were adjacent to the last column, that is, as if the spectrogram formed a cylinder.
[0106] Before the or each neuron layer, a sliding window can be applied to the spectrogram of directions of arrival calculated by the first calculation module 35 .
[0107] Therefore, the spectrogram of the arrival direction is processed into an image by the first neural network 75.
[0108] The first neural network 75 provides first data as its output.
[0109] The second neural network 80 is a fully connected type neural network. The second neural network 80 can take as input a vector including the normalized energy level of each microphone 25 calculated by the second calculation module 40 .
[0110] The second neural network 80 provides second data as its output.
[0111] The concatenation block 83 is connected to the outputs of the first neural network 75 and the second neural network 80 and is capable of concatenating the first and second data into a single vector to form concatenated data.
[0112] The third neural network 85 is connected to the cascade block 83. The third neural network 85 is, for example, of a fully connected type.
[0113] The third neural network 85 is preferably capable of classifying the concatenated data from a plurality of categories (eg, nine categories).
[0114] For example, all but one of the categories correspond to predefined positions of the wall 15 relative to the loudspeaker 10, and the last category corresponds to the absence of the wall 15. The position of the wall 15 is defined here as the direction in which the wall 15 is positioned relative to the loudspeaker 10 in a predefined reference system centered at the loudspeaker 10. Thus, each category corresponds to a predefined position of the wall 15, which is defined by the angular interval at which the wall 15 is positioned relative to the loudspeaker 10.
[0115] If the plurality of categories includes nine categories, each of the first eight categories corresponds to an angular interval of 45°, and the ninth category corresponds to the absence of the wall 15 .
[0116] To this end, the third neural network 85 is able to provide an output vector containing nine components at its output. The activation function of the neurons in the last layer of the third neural network 85 is preferably a SoftMax function. Therefore, the value of each component of the output vector is between 0 and 1, and the sum of the values of all components of the vector is equal to 1. The value of each component is then the probability of belonging to the corresponding category.
[0117] For example, if the component with the highest value in the output vector is a component associated with a category corresponding to the absence of the wall 15, the determination module 45 can determine the absence of the wall 15. Otherwise, the determination module 45 can then determine the presence of the wall 15.
[0118] As an optional addition, the determination module 45 can also determine the position of the wall 15 as the position defined by the class whose output component has the highest value in the output vector.
[0119] Optionally, the determination module 45 is configured to compare the highest value with a predefined threshold value. If the highest value is lower than the predefined threshold value, the determination module 45 can determine that the detection of the wall 15 is uncertain, and instruct the first calculation module 35 and the second calculation module 40 to repeat the calculation based on the second audio signal acquired later.
[0120] The adaptation module 50 can receive the position of the wall 15 or information on the absence of the wall 15 from the determination module 45 .
[0121] The adaptation module 50 is able to reconfigure the broadcasting mode of the loudspeaker 10 based on the determination of the presence or absence of the wall 15 and, more preferably, based on the location of said wall 15 (if applicable).
[0122] exist Figure 1 In the example of FIG. 1 , a diffusion pattern is associated with each of the following positions of the wall 15: facing one loudspeaker 20, facing the other loudspeaker 20, 90° to a line connecting the two loudspeakers 20 in one direction, and 90° to a line connecting the two loudspeakers 20 in the opposite direction. Thus, these positions define four broadcast patterns.
[0123] For each of the four broadcast modes, the control device 27 applies a delay to one of the loudspeakers 20 , for example to compensate for the presence of the wall 15 .
[0124] The fifth broadcast mode is defined as a neutral broadcast mode, in which the loudspeakers 20 are controlled synchronously, ie without delay. The adaptation module 50 can, for example, associate the fifth broadcast mode with all other positions and with the determination that the wall 15 is not present.
[0125] In the fifth broadcast mode, the different channels are preferably balanced in phase and amplitude.
[0126] In the first, second, third and fourth broadcast modes, further processing is preferably performed to further improve the sound rendering for users located opposite the wall relative to the loudspeaker 10. For example, the processing includes phase and amplitude equalization, and decomposing the signal into components to be propagated to the front and rear of the loudspeaker 10.
[0127] The adaptation module 50 is adapted to send the reconfigured broadcast mode to the control device 27 so that the control device 27 controls the loudspeaker 20 according to the reconfigured broadcast mode.
[0128] Now refer to Figure 5 Describing the operation of the speaker 10, Figure 5 A flow chart of a method for determining the presence or absence of a wall 15 is shown.
[0129] During a transmission step 110, the control device 27 receives audio content from a source and sends an activation control to the loudspeaker 20 to transmit the audio content.
[0130] At some point, the accelerometer 28 detects that the loudspeaker has become stationary after having been moved. The accelerometer then sends an instruction to the detection device 30 indicating that the loudspeaker 10 has become stationary. The audio content emitted by the loudspeaker 20 from this moment on is the emitted audio signal.
[0131] During an acquisition step 120 , each microphone 25 acquires a received audio signal.
[0132] Then, during a first calculation step 130 , the first calculation module 35 calculates a spectrogram of the directions of arrival of the audio signal components received by the loudspeaker 10 from the received audio signals acquired by each microphone 25 , for example as described above.
[0133] During a second calculation step 140 , the second calculation module 40 calculates the energy level of the received audio signal acquired from each microphone 25 , for example as previously described.
[0134] Preferably, the first calculation step 130 and the second calculation step 140 are performed simultaneously.
[0135] Then, during a determination step 150 , the determination module 45 determines the presence or absence of a wall 15 in the environment of the loudspeaker 10 by applying the neural network model 70 to the spectrogram of the directions of arrival and the calculated energy levels.
[0136] For example, the determination module 45 applies the first neural network 75 to the spectrogram to obtain first data, and applies the second neural network 80 to the energy level to obtain second data. The determination module 45 then applies the concatenation block 83 to the first and second data to form concatenated data. The determination module 45 then provides the concatenated data to the third neural network 85 to determine the output vector.
[0137] According to this optional addition, the determination module 45 determines the position of the wall 15 relative to the loudspeaker 10 or the absence of the wall 15 from the output vector, as described above.
[0138] Optionally, during the adaptation step 160 , the adaptation module 50 reconfigures the broadcast mode of the loudspeaker 10 according to the position of the wall 15 relative to the loudspeaker 10 or the absence of the wall 15 , for example by sending the reconfigured broadcast mode to the control device 27 .
[0139] Using the speaker 10 according to the present application, the presence or absence of the wall 15 can be accurately detected.
Claims
1. A loudspeaker (10), comprising: - at least one loudspeaker (20) capable of emitting a transmitted audio signal, - a plurality of microphones (25), each of which is capable of acquiring a received audio signal, If a wall (15) is present in the environment of the loudspeaker (10), the received audio signal contains the transmitted audio signal which has been reflected by the wall (15), and - electronic detection means (30) of the wall (15), connected to each microphone, The electronic detection device comprises: a first calculation module (35) capable of calculating a spectrogram of the direction of arrival of the audio signal received by the loudspeaker (10) based on the received audio signal acquired by each microphone (25), It is characterized in that The electronic detection device also comprises: a second calculation module (40) capable of calculating the energy level of the received audio signal acquired from each microphone (25), and A determination module (45) is capable of determining the presence of the wall (15) or the absence of the wall (15) in the environment of the loudspeaker (10) by applying a neural network model (70) to the spectrogram of the arrival direction and the determined energy level.
2. The loudspeaker (10) according to claim 1, wherein the neural network model (70) comprises: a first convolutional neural network (75) capable of receiving a frequency spectrum of an arrival direction, a second neural network (80) capable of receiving the energy level, a cascade block (80) capable of cascading data from the first convolutional neural network (75) and the second neural network (80) to form cascaded data, and a third neural network (85) capable of processing the cascaded data to determine the presence or absence of the wall (15).
3. A loudspeaker (10) according to claim 2, wherein the first neural network (75) comprises a circular sliding window and at least one neural layer, wherein the sliding window can be applied to a spectrogram of the direction of arrival before the neural layer.
4. The speaker (10) according to claim 1, wherein if the wall (15) exists in the speaker (10) environment, the neural network model (70) is capable of determining the position of the wall (15) relative to the speaker (10) from a plurality of predefined positions based on the frequency spectrum of the arrival direction and the calculated energy level.
5. The loudspeaker (10) according to claim 1, wherein the at least one loudspeaker (20) is capable of emitting the emitted audio signal according to a broadcast mode of the loudspeaker (10), The electronic detection device (30) further comprises an adaptation module (50) which is capable of reconfiguring the broadcasting mode of the loudspeaker (10) according to determining the presence or absence of the wall (15).
6. A loudspeaker (10) according to claim 1, wherein the frequency spectrum of the arrival direction is a matrix containing a plurality of values, each value corresponding to the power of the audio signal received by the loudspeaker (10) according to a predefined angular interval relative to a predefined reference system centered on the center of the loudspeaker (10), and according to a predefined frequency interval.
7. The speaker (10) of claim 1, wherein the transmitted audio signal is included in an audio stream broadcasting instructions from a user.
8. The speaker (10) according to any one of claims 1-7, further comprising: an accelerometer (28) capable of determining whether the speaker (10) is stationary and transmitting a calculation instruction when the speaker (10) is stationary, and a first (35) calculation module and a second (40) calculation module capable of calculating a frequency spectrum and an energy level of an arrival direction after receiving the calculation instruction from the accelerometer (28).
9. A method for detecting a wall (15) in an environment of a loudspeaker (10), the loudspeaker (10) comprising: at least one loudspeaker (20), a plurality of microphones (25) and an electronic device (30) connected to each microphone (25) for detecting the wall (15), The method comprises the following steps: - transmitting (110) an audio signal transmitted by the at least one loudspeaker (20), - acquiring (120) the audio signal received by each microphone (25), If the wall (15) exists in the environment of the speaker (10), the received audio signal includes an audio signal that has been reflected on the wall (15), - calculating (130) a spectrogram of the direction of arrival of the audio signal received by the loudspeaker (10) based on the received audio signal acquired by each microphone (25), Characterized in that the method further comprises the following steps: - calculating (140) the energy level of the received audio signal acquired from each microphone (25), and - determining the presence of said wall (15) or the absence of said wall (15) in the environment of the loudspeaker (10) by applying a neural network model (70) to the spectrogram of the directions of arrival and the determined energy levels.
10. A computer program product comprising software instructions, wherein when executed by a computer, the software instructions implement the detection method of claim 9.