• 专利附录
  • 专利说明
  • 权利要求
  • 类似技术
  • 同族专利
  • 引用文献
  • 法律状态
  • 优先权
    • 专利摘要:
    method and device to automatically detect marine animals. a method for automatically detecting marine animals, conducted by a detection device, comprises: - a step of obtaining the acoustic signal measurements (1) collected by at least one acoustic sensor in an underwater environment; - at least one of a first branch (3) to detect frequency-modulated sounds and a second branch (4) to detect impulsive sounds; each branch comprising a step of detecting sounds by: deploying several detection channels in parallel, each having a different and fixed value for at least a degree of freedom; select the detection channel having a maximum signal to noise ratio; and comparing the signal to the noise ratio of the selected detection channel at a given limit; - a step (32, 42, 5) of making an alarm decision, indicating the presence of at least one marine animal, as a function of an exit from the first branch and / or an exit from the second branch.
    公开号:BR112014000007B1
    申请号:R112014000007-7
    申请日:2012-06-19
    公开日:2020-12-15
    发明作者:Cédric Gervaise
    申请人:Sercel；
    IPC主号:
    专利说明:

    1. FIELD OF THE INVENTION
    The field of the invention is that of signal processing and submerged acoustics.
    More specifically, the invention refers to a technique to automatically detect marine animals, ie, a technique that allows the detection of the presence of marine animals from the detection of their sound emissions, using a PAD system (“Passive Acoustic Detection ”).
    The present invention can be applied notably, but not exclusively, to detect the presence of marine mammals, including cetaceans (mystics and odontocetes).
    For marine mammals, sound production is divided into two families: communication signals (frequency modulated signals, whistles) and signals used to perceive their environment (pulses, clicks). For example, all cetacean sound production includes Very Low Frequency emissions from mysticisms, Medium Frequency emissions from odontocetes (whistles) and High Frequency emissions from odontocetes (clicks).
    It should be noted that if a PAD system allows you to hear and detect signals of biological origin, it can also hear and detect any sound contribution parasites created by other activities.
    A wide spectrum of applications of the invention is identified from the needs of scientists working in the field of marine ecology who want to hold marine animal observation tools, to operators in the sea world who wish to limit the negative interactions of their activity on marine animals ( fishing activity, military activity with sonar emissions, acquisition of geophysical data to analyze the seabed (eg, oil prospecting industry using the seismic method), etc.). 2. TECHNOLOGICAL HISTORY
    Corresponding to a strong need and supported by an active scientific community, numerous algorithmic solutions have been proposed for the detection of marine animals.
    A series of international workshops held since 2003 echoes this dynamic: • Special Edition, “Detection and localization of marine mammals using passive acoustics”, Canadian Acoustics, Vol. 32, 2004. • Special Edition, “Detection and localization of marine mammals using acoustics” passive ”, Applied Acoustics, vol. 67, 2006. • Special Edition, “Marine mammal detection and classification using passive acoustics”, Canadian Acoustics, Vol. 36, 2008. • Special Edition, “Marine mammal detection, classification, location and census with passive acoustic monitoring” , Applied Acoustics, vol. 71, 2010.
    Through these four references, algorithmic solutions for the detection of clicks can be identified: a) Using the classic energy descriptor solution: W. Mr. X. Zimmer, J. Harwood, PL Tyack, P. Johnson, and PT Madsen, “Passive acoustic detection of deep-sea beaked whales”, The Journal of the Acoustical Society of America, vol.124, pp. 2823-2832, 2008. b) Using the original solution from Teager's descriptor: V. Kandia and Y. Stylianou, “Detection of sperm whale clicks based on the Teager-Kaiser energy operator”, Applied Acoustic, Vol. 67 , pp. 1144-1163, 2006. c) Using the original kurtosis descriptor solution: C. Gervaise, A. Barazzutti, S. Busson, Y. Simard, and N. Roy, “Automatic detection of Bioacoustic impulses based on kurtosis under phrase signal for noise ratio ”, Applied Acoustics, vol. 71, pp. 1020-1026, 2010.
    For the detection of whistles, algorithmic solutions were proposed: a) Using the spectrogram: DK Mellinger and CW Clark, “Recognizing transient low-frequency whale sounds by spectrogram correlation”, The Journal of the Acoustical Society of America, Vol. 107, pp. 3518-3529, 2000. b) Using the Hilbert Huang Transformation: Adam O (2006), “Advantages of Hilbert Huang transformation for marine mammal signal analysis”, J. Acoust. Soc. Am 120: 2965-2973. c) Using the ambiguity function in high orders and deformation operators: C. loana, C. Gervaise, Y. Stephan, and JI March, “Analysis of underwater mammal vocalizations using time frequency phase tracker”, Applied Acoustics, vol. 71, pp. 1070-1080, 2010.
    In general, at best, the solutions known above have a frequency adaptability (allowing to select the frequency band in which the signal exists, and to reject ambient noise in other frequency bands) and they select a specific detection test , which is compared to an estimated value of this detection test in the case of only a measurement noise.
    Unfortunately, these known solutions suffer from several limitations: • They cannot be fitted into an autonomous communication system; • Their performances are fixed or depend on the presence of a trained operator to adjust the settings or the architecture of the algorithmic solutions; • They do not automatically adapt to the properties (which are often variable) of biological sound production; • They do not handle all the sound production of marine animals in a single process; • They do not include learning and rejecting false alarms generated by ambient noise. 3. OBJECTIVES OF THE INVENTION
    The invention, in at least one embodiment, is aimed notably at overcoming these different disadvantages of the prior art.
    More specifically, it is a goal of at least one embodiment of the invention to provide a technique for detecting marine animals, that technique treating sound production of different marine animals in a single process (eg, all sound production of marine mammals, including cetaceans (mystics and odontocetes)).
    It is also a goal of at least one embodiment of the invention to provide a technique for detecting marine animals, that technique having the ability to adapt the properties of biological sound sources and ambient noise properties.
    It is another goal of at least one embodiment of the invention to provide a technique for detecting marine animals, this technique being performed automatically, without requiring an operator and with a minimum number of preliminary configuration settings.
    It is an additional goal of at least one embodiment of the invention to provide a technique for detecting marine animals, allowing real-time operation in an embedded system (i.e., an autonomous communication system).
    It is an additional goal of at least one embodiment of the invention to provide a technique for detecting marine animals, allowing to identify abiotic sound detections (eg, due to the seismic mechanism) and to reject them. 4. SUMMARY OF THE INVENTION
    A specific embodiment of the invention proposes a method for automatically detecting marine animals, which is conducted by a detection device and comprises: - a step of obtaining the measurements of the acoustic signal collected by at least one acoustic sensor in an underwater environment; - at least one of the following branches: * a first branch comprising a step of detecting sounds modulated by frequency by: implanting in several first detection channels in parallel, each having a different and fixed value for at least a degree of freedom; select the first detection channel having a maximum signal for noise ratio; and comparing the signal to the noise ratio of the first selected detection channel to a first determined limit; * a second branch comprising a step of detecting impulsive sounds by: implanting in several seconds detection channels in parallel, each having a different value and fixed by at least a degree of freedom; selecting the second detection channel having a maximum signal to noise ratio; and comparing the signal to the noise ratio of the selected second detection channel to a second determined limit; - a step of making an alarm decision, indicating the presence of at least one marine animal, as a function of an exit from the first branch and / or an exit from the second branch.
    In this way, an innovative method is proposed to process the measurements (sound pressure levels) collected by at least one acoustic sensor, with the purpose of producing a warning (alarm decision) that informs the presence of one or several marine animals in the detection variation of the acoustic sensor.
    Each branch having several detection channels in parallel, this technique has numerous advantages: it has the ability to adapt the properties of biological sound sources and the properties of ambient noise, and can be performed automatically (without requiring an operator and with a number minimum of preliminary configuration settings).
    When the two branches are used (one for processing frequency-modulated sounds and the other for processing impulsive sounds), this technique allows to treat the sound productions of different marine animals in a single process.
    According to a specific resource, the said at least one degree of freedom belongs to the group comprising: - methods, and corresponding parameters, to present the acoustic signal measurements in a new representation space; - quantitative signal resources in which the acoustic signal measurements are mapped in the new representation space; - methods to estimate the noise characteristic.
    In this way, the parallel detection channels of each branch can conduct different detections and the quality of the alarm decision is improved.
    According to a specific resource, each of the first detection channels uses a fast Fourier transformation, with a different length, to present the acoustic signal measurements in a new representation space, and uses energy as a quantitative signal resource. in which the acoustic signal measurements are mapped in the new representation space.
    This improves the processing of frequency-modulated sounds.
    According to a specific feature, each second detection channel implements the following steps: - use a passband filter, with a different passband, to present the acoustic signal measurements in a new representation space; - compute the first and second signals to the noise ratios using respectively a first and a second quantitative signal resource in which the acoustic signal measurements are mapped in the new representation space, said first and second quantitative signal resources being associated for different order statistics; and - selecting a maximum ratio between the first and second signals to the noise ratios, to be used in the step of selecting the second detection channel having a maximum signal to the noise ratio.
    This improves the processing of impulsive sounds.
    According to a specific resource, said first quantitative signal resource is energy associated with a second order statistic, and said second quantitative signal resource is kurtosis associated with a fourth order statistic.
    This further improves the processing of impulsive sounds.
    According to a specific resource, at least one of the said branches (first and second) comprises: - a learning step, adapted to determine the rejected points of a time frequency grid, as a function of a plurality of sounds detected effectively successive; and - a reject step, adapted to reject an effectively detected sound that is located, in the time frequency grid, in one of the rejected points; and the step of making an alarm decision is conducted as a function of the sound or sounds actually detected and not rejected.
    This makes it possible to identify the abiotic sound detections (eg, due to the seismic mechanism) and to reject them.
    According to a specific resource, said learning step comprises the following steps, for at least a certain point in the time frequency grid: - obtaining countless effectively detected sounds that are mapped at the given point in the time frequency grid, between a plurality of successively detected sounds in a specified number of iterations; - decide that a given point in the time frequency grid is a rejected point if that number is higher than a given limit. This improves the learning stage.
    According to a specific resource, the method is deployed in real time on said detection device, and comprises a step of transmitting said alarm decision to a remote management device.
    In another embodiment, the invention relates to a computer program product comprising the program code instructions for implementing the aforementioned method (in any of its different embodiments) when said program is executed on a computer or a processor.
    In another embodiment, the invention relates to a non-transitory, computer-readable transport medium, storing a program that, when executed by a computer or a processor, causes the computer or processor to conduct the aforementioned method (in any of its different achievements).
    A specific embodiment of the invention proposes a detection device to automatically detect marine animals, comprising: - means for obtaining the acoustic signal measurements collected by at least one acoustic sensor in an underwater environment; - at least one of the following processing medium: * a first processing medium, allowing the detection of frequency-modulated sounds, and comprising: first diverse parallel medium for detection, each having a different and fixed value for at least a degree of freedom ; means for selecting the first means for detection having a maximum signal to noise ratio; and means for comparing the signal to the noise ratio of the first medium selected for detection to a first determined limit; * a second means of processing, allowing the detection of impulsive sounds, and comprising: second means in parallel for detection, each having a different and fixed value for at least a degree of freedom; means for selecting the second means for detection having a maximum signal to noise ratio; and means for comparing the signal to the noise ratio of the second means selected for detection to a second determined limit; - means for making an alarm decision, indicating the presence of at least one marine animal, as a function of an output from the first processing medium and / or an output from the second processing medium. 5. LIST OF FIGURES »
    Other features and advantages of the realizations of the invention will be apparent from the following description, provided by means of indicative and non-exhaustive examples and from the accompanying drawings, in which: - Figure 1 is a schematic illustration of the detection method according to a specific embodiment of the invention, comprising a branch for frequency-modulated sounds and a branch for impulsive sounds; Figure 2 is a generic illustration of a detection block comprised in each of the branches shown in Figure 1, and itself comprising a detection stage and a learning and rejection stage; Figure 3 is a generic illustration of the detection stage shown in Figure 2; Figure 4 is a schematic illustration of a specific realization of the detection stage comprised in the branch for the frequency modulated sounds; Figure 5 is a schematic illustration of a specific realization of the detection stage comprised in the branch for impulsive sounds; - Figure 6 is a generic illustration of the learning and rejection stage shown in Figure 2; - Figure 7 is a schematic illustration of a specific realization of the learning and rejection stage shown in Figure 2; - Figure 8 shows the simplified structure of a detection device according to a specific embodiment of the invention. 6. DETAILED DESCRIPTION.
    In all figures in this document, identical elements and steps are designated by at least a numerical reference sign.
    In the example described below, we consider the detection of the presence of marine mammals, including cetaceans (mystics and odontocetes).
    Referring now to Figure 1, we present a detection method according to a specific embodiment of the invention.
    Input 1 consists of measurements (sound pressure levels) collected by an acoustic sensor (eg, a pressure sensor (hydrophone) or a particle motion sensor (geophone, accelerometer)) embedded, for example, in a standard instrumentation for geophysical research (air gun, streamers, ...). As shown in Figure 4, the measurements collected by the acoustic sensor 11, before being supplied to two branches 3 and 4, are, for example, amplified by a preamplifier 12 with variable gain, and then converted to digital form by a converter analog / digital (ADC) 13.
    Output 2 of the diagram is a warning that informs you of the presence of some marine mammals in the detection variation of the acoustic sensor.
    The input signal (measurements) is processed per frame of T seconds, so that the detection and warning have the sample rate of T seconds.
    If the method is implemented in real time on a detection device, it comprises a step of transmitting the final alarm decision to a remote management device.
    The process is divided into two branches, one (reference 3) for the processing of frequency modulated sounds (vocalization of mysticisms, whistles of odontocetes) and the other (reference 4) for the processing of impulsive sounds (clicks of odontocetes).
    Each branch (3 or 4) comprises a detection block (reference 31 or 41, and detailed below) for detection and a warning block (reference 32 or 42).
    The warning block with reference 32 takes a first intermediate alarm decision 33, indicating the presence of at least one marine mammal, as a function of the frequency-modulated sound or sounds 34 actually detected by the reference block 31.
    The warning block with reference 42 takes a second intermediate alarm decision 43, indicating the presence of at least one marine mammal, as a function of the sound or impulsive sounds 44 effectively detected by the detection block 41.
    The final alarm decision (i.e., output 2) is a function (logical function “OR” in this example) of the first and second intermediate alarm decisions 33 and 43.
    The operation of each warning block 32, 42 can be summarized as follows. As a first step, the warning block incorporates several individual detected sounds (for warning block 32: whistles from detection block 31; for warning block 42: clicks from detection block 41) into a single detection metric. The detection metric can be defined by the operator, for example: the proportion of time covered by whistles or clicks in a reference travel time interval; or the number of events (whistles or clicks) detected in a reference progress time range. Then, the time series of the detection metric is compared to a limit (rigid or flexible). Every time the metric is more than the threshold, a detection warning is declared.
    As shown in Figure 2. each of the detection blocks 31 and 41 comprises a detection stage 6 (also called “stage 1”), for sound detection (see below the description of Figures 3, 4 and 5), and a learning and rejection stage 7 (also referred to as “stage 2” below), for learning and abiotic sound rejection (see description of Figures 6 and 7 below). The output of detection stage 6 is with reference 35 (for branch with reference 3) or 45 (for branch with reference 4). The output of the learning and rejection stage 7 is the output of the detection stage 6 and, as already mentioned above, it is with reference 34 (for branch with reference 3) or 44 (for branch with reference 4).
    Figure 3 is a generic illustration of detection stage 6 shown in Figure 2. Basically, it comprises: • The N detection channels in parallel, with reference 61i to 61N (also called “substage i (regulation i)”, with i an integer from 1 to N and N an integer greater than 1); • A block 62 to dynamically select at each moment (i.e., for each period T) the detection channel that optimizes the SNR (Signal to Noise Ratio) output; and • A block 63 to perform detection only on the selected channel, when comparing the SNR of the selected channel to a determined limit 64 (computed from the theoretical assumption and the maximum acceptable false alarm rate).
    Each of the N detection channels consists of a first operation to present measurements in a new representation space (basically, time and frequency) and, in this new representation space, a quantitative resource of sounds is mapped. Both the representation spaces and the signal resource are chosen to separate useful signals from noise as well as possible. Then, in this new representation space, the detection is performed by estimating a local SNR where the Noise is estimated from the measurement maps.
    In order to adapt to a wide range of useful signal waveforms and noise characteristics, each of the N detection channels has a different and fixed value for one or several degrees of freedom: • The method and its settings (parameters) θ to change the representation space (ie, present the acoustic signal measurements in a new representation space), • The quantitative signal feature in which the processed data is mapped, in the new representation space, and • O method to estimate the noise characteristic.
    Once these degrees of freedom are chosen, each detection channel is optimized for a useful signal type and a noise type.
    It should be kept in mind that the SNR value well before the final detection (i.e., the output of block 62 and the input of block 63) is an indicator of optimality of choice for degrees of freedom.
    This innovative general scheme, detailed above with Figures 1 and 2, can be applied and implemented, for example, in a DSP (Digital Signal Processor) for the two channels of our detector.
    More generally, it is possible to consider that the general innovative scheme can be implemented equally well: • By executing a set of computer instructions executed by a reprogrammable computing machine, such as a PC-type mechanism, a DSP or a microcontroller; or other • By a dedicated hardware machine or component, such as an FPGA (Field Programmable Port Arrangement), an ASIC (Application Specific Integrated Circuit) or any other hardware module.
    In the event that the solution is deployed on a reprogrammable computing machine, the corresponding program (ie, the instruction set) can be stored on a non-transitable, computer-readable transport medium that is separable (for example, a floppy disk, a CD-ROM or a DVD-ROM) or not separable. Figure 8 shows the simplified structure of a detection device according to a specific embodiment of the invention, with implantation in a DSP. Device 85 for automatically detecting marine animals comprises a DSP 81, a read-only memory (ROM) 82 and a random access memory (RAM) 83. Read-only memory 82 stores the executable code of programs, which, when they are executed by the DSP, it allows the implementation of the innovative general scheme detailed above with Figures 1 and 2. Upon initialization, the program code instructions mentioned above are transferred to random access memory 83 in order to be executed by DSP 81. Random access memory 83, likewise, includes records to store the variables and parameters required for this execution. The DSP 81 receives measurements (sound pressure levels) (i.e., input 1) and delivers the final alarm decision (i.e., output 2). Figure 4 is a schematic illustration of a specific realization ("multi-FFT" approach) of detection stage 100 comprised in branch 3 for frequency modulated sounds.
    In this embodiment, the detection stage 100 comprises: • A low-pass filter 101; • A decimator 102 with a decimator factor of 4; • In parallel, two (N = 2) detection channels 1031 and 1032, whose FFT lengths are 512 and 2048, respectively. Clearly, the number N can be greater than 2; • A block 104 to dynamically select each time (i.e., for each T period) the detection channel that optimizes the SNR (Signal to Noise Ratio) output; • A block 106 to perform the detection only on the selected channel, when comparing the SNR of the selected channel to a determined limit.
    As mentioned above, if the method is implemented in real time on a detection device, it comprises a step of transmitting the final alarm decision to a remote management device. For example, in a context of acquisition of seismic data in a marine environment, the detection device can be understood in a coil (the sensors are distributed along cables in order to form linear acoustic antennas commonly called “coils” or “Seismic coils”; the seismic coil network is towed by a seismic vessel). With this real-time signal processing restriction, it is not possible to use a very long FFT length, which would involve a very long delivery time from the detection device to the management device. The pair of lengths of FFT 512 and 2048 allows to meet the restriction of signal processing in real time.
    Each of the N detection channels of the detection stage is defined as follows: • The method of changing the representation space is a Fast Fourier Transformation (FFT) with the length L of the FFT chosen as parameters to adjust; • The signal resource (in which the acoustic signal measurements are mapped in the new representation space) is energy; and • The noise estimation method is an order low pass Infinite Impulse Response filter.
    In an alternative embodiment, the method for changing the representation space is not a Fast Fourier Transformation (FFT), but a time frequency map method between the following list (not exhaustive): zero crossing, empirical decomposition ( EMD), filter bank or small wave transformation.
    Figure 5 is a schematic illustration of a specific realization ("filter bank with multiple order statistics" approach) of detection stage 200 comprised in branch 4 for impulsive sounds.
    In this embodiment, the detection stage 200 comprises: • A high-pass filter 201; • A bank 202 of N (eg, N = 6) passband filters running in parallel, to present the acoustic signal measurements in a new representation space; • In parallel, N detection channels 2031 to 203N. Each detection channel computes the first and second SNRs (signal to noise ratios) using respectively a first and a second quantitative signal resource in which the acoustic signal measurements are mapped in the new representation space. The maximum between the first and second SNRs is selected for the detection channel in question. The first and second quantitative signal resources are associated with different order statistics. In a preferred embodiment, the first quantitative signal resource is energy associated with a second order statistic, and the second quantitative signal resource is kurtosis associated with a fourth order statistic. The noise estimation method is an order low-pass Infinite Impulse Response filter. In alternative embodiments, the pair (energy kurtosis) can be changed in any other pair comprising two items from the following list (not exhaustive): energy, obliquity, kurtosis and alpha state parameter. • A block 204 to dynamically select at each moment (i.e., for each period T), the detection channel that optimizes the SNR (Signal to Noise Ratio) output; • A 206 block to perform detection on the selected channel only, when comparing the SNR of the selected channel to a specified limit.
    With the description of Figures 6 and 7. a specific realization of the learning and rejection stage 7 shown in Figure 2.
    Detection stage 100 (“multi-FFT approach” in Figure 4), comprised in branch 3 for frequency-modulated sounds, and detection stage 200 (“filter bank with multiple order statistics” in Figure 5 ), included in branch 4 for impulsive sounds, to optimize the detection rate of any sound measured by the acoustic sensor, whatever the nature of the source.
    However, a passive detector device (eg, embedded in standard geophysical instrumentation) will encounter several abiotic sound productions generated by the instrumentation itself. Its detection is a real sound detection, but a false detection of marine mammals.
    In order to reduce the false alarm rate due to abiotic sounds, the learning and rejection stage 7 is added, in each of branches 3 and 4, to learn what these abiotic sound detections are and how to eliminate them. The learning and rejection stage 7 comprises: • A learning block 71, adapted to determine the rejected points of a time frequency grid, as a function of a plurality of effectively detected sounds; and • A rejection block 72, adapted to reject an effectively detected sound that is located, in the time frequency grid, in one of the rejected points.
    Learning block 71 uses the realistic assumption that abiotic sound detections are common, whereas marine mammalian detections are scarce. A running time frame period of length T is defined. In this period T, the sound detection stage 6 (ie, the detection stage 100 for the branch 3 processing the frequency modulated sounds, and the detection stage 200 for the branch 4 processing the impulsive sounds) maps the signal in a time frequency grid. Let G (t, f) be that grid. When the time in progress, the number of detections for each grid point G (t, f) is added with a time memory of D and divided by the total number of iterations it contains in D, to obtain the probability that a point of the grid is a false alarm due to abiotic sounds. Learning block 71 decides that such a point in the time frequency grid is a rejected point if the associated probability is higher than a given limit.
    In grid 73 shown at the top of Figure 7, the size of the circle located at a given point on the grid is proportional to the probability that that particular point is a false alarm due to abiotic sounds (ie, the size of the circle is proportional to the number of effectively detected sounds that are located at that grid point).
    In grid 74 shown at the bottom of Figure 7, the grid points marked with a cross are the rejected points, to be used by the rejection block 72 to 5 to decide whether an effectively detected sound should be rejected or not.
    权利要求:
    Claims (11)
    [0001]
    1. Method for automatically detecting marine animals, characterized by the fact that it is conducted by a detection device (85) and comprises: - a step of obtaining acoustic signal measurements (1) collected by at least one acoustic sensor in an underwater environment ; - at least one of the following branches: - a first branch (3) comprising a step (31) of detecting frequency modulated sounds by: deploying several first detection channels in parallel (1031-1032); selecting (104) the first detection channel having a maximum signal to noise ratio; and comparing (106) the signal to the noise ratio of the first selected detection channel to a first determined limit; - a second branch (4) comprising a step (41) of detecting impulsive sounds by: implanting several second detection channels in parallel (2031-203N); selecting (204) the second detection channel having a maximum signal at noise ratio; and comparing (206) the signal to the noise ratio of the selected second detection channel to a second determined limit; - a step (32, 42, 5) of making an alarm decision, indicating the presence of at least one marine animal, as a function of an exit from the first branch and / or an exit from the second branch; in which each first or second detection channel comprises: - presenting acoustic signal measurements in a new representation space, said new representation space providing a representation of time and frequency; - in the referred new representation space, estimate a signal to noise ratio (SNR1, SNRi, SNRN); wherein each first or second detection channel has a different and fixed value for at least one degree of freedom belonging to the group comprising: - methods, and corresponding parameters, for presenting the acoustic signal measurements in said new representation space; - quantitative signal resources in which the acoustic signal measurements are mapped in the new representation space; - methods to estimate the noise characteristic.
    [0002]
    Method according to claim 1, characterized in that it comprises said first branch and said second branch.
    [0003]
    Method according to any one of claims 1 to 2, characterized by the fact that each of the first detection channels uses a fast Fourier transformation, with a different length, to present the acoustic signal measurements in a new space. representation, and uses energy as a quantitative signal resource in which the acoustic signal measurements are mapped in the new representation space.
    [0004]
    Method according to any one of claims 1 to 3, characterized in that each second detection channel implements the following steps: - using a passband filter, with a different passband, to present the measurements of acoustic signal in a new representation space; - compute the first and second signals to the noise ratios using respectively a first and a second quantitative signal resource in which the acoustic signal measurements are mapped in the new representation space, the referred and second quantitative signal resources being associated with different statistics of order; and - selecting a maximum ratio between the first and second signals to the noise ratios, to be used in the step of selecting the second detection channel having a maximum signal to the noise ratio.
    [0005]
    5. Method according to claim 4, characterized by the fact that said first quantitative signal resource is energy associated with a second order statistic, and said second quantitative signal resource is short if associated with a fourth order statistic .
    [0006]
    Method according to any one of claims 1 to 5, characterized in that at least one of said first and second branches comprises: - a learning step (71), adapted to determine the rejected points of a frequency grid of time, as a function of a plurality of detected sounds effectively successive; and - a rejection step (72), adapted to reject an effectively detected sound that is located, in the time frequency grid, in one of the rejected points; and so that the step of making an alarm decision is conducted as a function of the sound or sounds actually detected and not rejected.
    [0007]
    7. Method according to claim 6, characterized by the fact that said learning step (71) comprises the following steps, for at least a certain point in the time frequency grid: - obtaining a number of effectively detected sounds that they are mapped at the given point in the time frequency grid, among a plurality of sounds detected effectively in succession over a determined number of iterations; - decide that the said point of the time frequency grid is a rejected point, if the said number is higher than a determined limit.
    [0008]
    Method according to any one of claims 1 to 7, characterized by the fact that it is implanted in real time in said detection device, and so that it comprises a step of transmitting said alarm decision to a remote management device .
    [0009]
    9. Computer program product characterized by the fact that it comprises the program code instructions for implementing the method according to at least one of claims 1 to 8, when said program is executed on a computer or a processor.
    [0010]
    10. Non-transitory, computer-readable transport media storing a program characterized by the fact that, when executed by a computer or a processor, it causes the computer or processor to conduct the method according to at least one of claims 1 to 8.
    [0011]
    11. Detection device (85) to automatically detect marine animals, characterized by the fact that it comprises: - means to obtain the acoustic signal measurements collected by at least one acoustic sensor in an underwater environment; - at least one of the following processing means: * a first processing means, allowing the detection of frequency-modulated sounds, and comprising: in parallel, some first means for detection; means for selecting the first means for detection having a maximum signal to noise ratio; and means for comparing the signal to the noise ratio of the first medium selected for detection at a first determined limit; * a second processing means, allowing the detection of impulsive sounds, and comprising: in parallel, a few second means for detection; means for selecting the second means for detection having a maximum signal to noise ratio; and means for comparing the signal to the noise ratio of the second medium selected for detection at a second determined limit; - means for making an alarm decision, indicating the presence of at least one marine animal, as a function of an output from the first processing medium and / or an output from the second processing medium in which each first or second detection medium comprises: - means for presenting acoustic signal measurements in a new representation space, said new representation space providing a representation of time and frequency; - means for estimating a signal to noise ratio (SNR1, SNRi, SNRN), in the aforementioned new representation space; where each first or second means for the detection channel has a different and fixed value for at least one degree of freedom belonging to the group comprising: - methods, and corresponding parameters, for presenting the acoustic signal measurements in the new representation space; - quantitative signal resources in which the acoustic signal measurements are mapped in the new representation space; - methods to estimate the noise characteristic.
    类似技术:
    公开号 | 公开日 | 专利标题
    BR112014000007B1|2020-12-15|METHOD AND DEVICE FOR AUTOMATICALLY DETECTING SEA ANIMALS
    Kyhn et al.2013|Clicking in a killer whale habitat: narrow-band, high-frequency biosonar clicks of harbour porpoise | and Dall’s porpoise |
    Nuuttila et al.2013|Identifying Foraging Behaviour of Wild Bottlenose Dolphins | and Harbour Porpoises | with Static Acoustic Dataloggers.
    Bahoura et al.2010|Blue whale calls classification using short-time Fourier and wavelet packet transforms and artificial neural network
    Jones et al.2008|Classification of broadband echoes from prey of a foraging Blainville’s beaked whale
    Širović et al.2009|Sounds of captive rockfishes
    Papale et al.2021|The higher, the closer, the better? Influence of sampling frequency and distance on the acoustic properties of short‐beaked common dolphins burst pulses in the Mediterranean Sea
    Wulandari et al.2016|Bioacoustic characteristic click sound and behaviour of male dolphins bottle nose |
    Holt et al.2019|Sounds associated with foraging and prey capture in individual fish-eating killer whales, Orcinus orca
    Lin et al.2018|Listening to the deep: Exploring marine soundscape variability by information retrieval techniques
    Mann et al.2016|Listening in the ocean: new discoveries and insights on marine life from autonomous passive acoustic recorders
    Macaulay2020|Passive acoustic monitoring of harbour porpoise behaviour, distribution and density in tidal rapid habitats
    Zhang et al.2015|Hearing characteristics of cephalopods: Modeling and environmental impact study
    Riera et al.2016|Effects of anthropogenic noise on fishes at the SGaan Kinghlas-Bowie Seamount Marine Protected Area
    Malinka et al.2021|Echolocation click parameters and biosonar behaviour of the dwarf sperm whale |
    Banlawe et al.2021|Optimal Frequency Characterization of Mango Pulp Weevil Mating Activity using Naïve Bayes Classifier Algorithm
    Harris et al.2019|Foraging Behavior and Disruption in Blue, Fin, and Humpback Whales in Relation to Sonar Exposure: The Challenges of Generalizing Responsiveness in Species with High Individual Variability.
    KR20210098197A|2021-08-10|Liquid attributes classifier using soundwaves based on machine learning and mobile phone
    Dähne et al.2020|No need to shout? Harbor porpoises | echolocate quietly in confined murky waters of the Wadden Sea
    Banlawe et al.2021|Decision Tree Learning Algorithm and Naïve Bayes Classifier Algorithm Comparative Classification for Mango Pulp Weevil Mating Activity
    Tanaka et al.2021|Automated Classification of Dugong Calls and Tonal Noise by Combining Contour and MFCC Features
    Isojunno et al.2021|Sperm whales exhibit variation in echolocation tactics with depth and sea state but not naval sonar exposures
    Liebschner et al.2016|Impacts of underwater noise on marine vertebrates: project introduction and first results
    Bolser et al.2022|Assessing the size spectra of marine fish communities with hydroacoustics: Examining the challenges of abundant schools, diverse assemblages, and variable orientations
    Hastings et al.2012|Marine bioacoustics and technology: The new world of marine acoustic ecology
    同族专利:
    公开号 | 公开日
    RU2593426C2|2016-08-10|
    CN103649778B|2017-02-15|
    BR112014000007A2|2017-07-18|
    CA2838060A1|2013-01-17|
    AU2012283315B2|2015-05-14|
    US20140293749A1|2014-10-02|
    RU2014102789A|2015-08-10|
    EP2546680A1|2013-01-16|
    WO2013007482A1|2013-01-17|
    CA2838060C|2021-01-19|
    CN103649778A|2014-03-19|
    EP2546680B1|2014-06-04|
    AU2012283315A1|2013-12-19|
    US9429666B2|2016-08-30|
    引用文献:
    公开号 | 申请日 | 公开日 | 申请人 | 专利标题
    US3893061A|1970-05-13|1975-07-01|Us Navy|Underwater acoustic detection system|
    US6181646B1|1997-01-07|2001-01-30|Hyroacoustics, Inc.|Geophysical exploration system using seismic vibrator source which provides a composite sweep|
    US6795374B2|2001-09-07|2004-09-21|Siemens Medical Solutions Usa, Inc.|Bias control of electrostatic transducers|
    JP3662921B1|2004-10-22|2005-06-22|千春 青山|Submarine resource exploration system and submarine resource exploration method|
    EP2546680B1|2011-07-13|2014-06-04|Sercel|Method and device for automatically detecting marine animals|EP2546680B1|2011-07-13|2014-06-04|Sercel|Method and device for automatically detecting marine animals|
    US20140358552A1|2013-05-31|2014-12-04|Cirrus Logic, Inc.|Low-power voice gate for device wake-up|
    US9557428B2|2014-03-12|2017-01-31|Sercel|Method for controlling quality of operation of a PAM system, corresponding computer program product, non-transitory computer-readable carrier medium and quality control device|
    US9625592B2|2014-03-12|2017-04-18|Sercel|Method for localizing a marine mammal in an underwater environment implemented by a PAM system, corresponding device, computer program product and non-transitory computer-readable carrier medium|
    US9823367B2|2015-06-16|2017-11-21|Sercel|Seismic shooting with mammal mitigation system and method|
    TWI567732B|2015-10-05|2017-01-21|Jin-Feng Lin|A Time-Frequency Energy Distribution Characteristic Analysis Method for Underwater Audio Signal Based on Empirical Mode Disassembly|
    BR112019025783A2|2017-06-08|2020-06-23|Total Sa|METHOD FOR PURCHASING A SET OF SEISMIC DATA IN A REGION OF INTEREST|
    WO2019002831A1|2017-06-27|2019-01-03|Cirrus Logic International Semiconductor Limited|Detection of replay attack|
    GB201713697D0|2017-06-28|2017-10-11|Cirrus Logic Int Semiconductor Ltd|Magnetic detection of replay attack|
    GB2563953A|2017-06-28|2019-01-02|Cirrus Logic Int Semiconductor Ltd|Detection of replay attack|
    GB201801528D0|2017-07-07|2018-03-14|Cirrus Logic Int Semiconductor Ltd|Method, apparatus and systems for biometric processes|
    GB201801530D0|2017-07-07|2018-03-14|Cirrus Logic Int Semiconductor Ltd|Methods, apparatus and systems for authentication|
    GB201801527D0|2017-07-07|2018-03-14|Cirrus Logic Int Semiconductor Ltd|Method, apparatus and systems for biometric processes|
    GB201803570D0|2017-10-13|2018-04-18|Cirrus Logic Int Semiconductor Ltd|Detection of replay attack|
    GB201801874D0|2017-10-13|2018-03-21|Cirrus Logic Int Semiconductor Ltd|Improving robustness of speech processing system against ultrasound and dolphin attacks|
    GB201801664D0|2017-10-13|2018-03-21|Cirrus Logic Int Semiconductor Ltd|Detection of liveness|
    GB2567503A|2017-10-13|2019-04-17|Cirrus Logic Int Semiconductor Ltd|Analysing speech signals|
    GB201801663D0|2017-10-13|2018-03-21|Cirrus Logic Int Semiconductor Ltd|Detection of liveness|
    GB201801659D0|2017-11-14|2018-03-21|Cirrus Logic Int Semiconductor Ltd|Detection of loudspeaker playback|
    US11264037B2|2018-01-23|2022-03-01|Cirrus Logic, Inc.|Speaker identification|
    US10362769B1|2018-03-23|2019-07-30|International Business Machines Corporation|System and method for detection of disease breakouts|
    US10915614B2|2018-08-31|2021-02-09|Cirrus Logic, Inc.|Biometric authentication|
    US11037574B2|2018-09-05|2021-06-15|Cirrus Logic, Inc.|Speaker recognition and speaker change detection|
    CN109541969B|2018-10-24|2021-04-23|深圳市亿兆互联技术有限公司|System and method for monitoring whales based on lora|
    FR3090131A1|2018-12-17|2020-06-19|Suez Eau France|Determination of the evolution of an underwater environment by acoustic analysis|
    法律状态:
    2017-02-07| B11Z| Dismissal: petition dismissal - article 216, par 2 of industrial property law|
    2017-03-14| B12F| Appeal: other appeals|
    2018-12-11| B06F| Objections, documents and/or translations needed after an examination request according art. 34 industrial property law|
    2019-12-10| B06U| Preliminary requirement: requests with searches performed by other patent offices: suspension of the patent application procedure|
    2020-10-20| B09A| Decision: intention to grant|
    2020-12-15| B16A| Patent or certificate of addition of invention granted|Free format text: PRAZO DE VALIDADE: 20 (VINTE) ANOS CONTADOS A PARTIR DE 19/06/2012, OBSERVADAS AS CONDICOES LEGAIS. |
    优先权:
    申请号 | 申请日 | 专利标题
    EP11305917.4A|EP2546680B1|2011-07-13|2011-07-13|Method and device for automatically detecting marine animals|
    EP11305917.4|2011-07-13|
    PCT/EP2012/061685|WO2013007482A1|2011-07-13|2012-06-19|Method and device for automatically detecting marine animals|
    [返回顶部]