Adaptive beamforming monitoring method and system for acoustic sensor arrays

Abstract

The present invention introduces a method and system for monitoring an area based on acoustic data acquired from at least one acoustic sensor array. The proposed method initiates by acquiring (10) acoustic data and environmental data including at least temperature. This data is processed (20) to generate a global acoustic image by employing a delay and sum based beamformer for adjusting the phase and amplitude of the acoustic data. The method then determines (25) the presence of objects of interest within the monitored area based on the global acoustic image. Upon identifying an object of interest, a location identification procedure is initiated (30). This procedure includes estimating the initial location of the object then refining the acquired acoustic data focusing on sections which are estimated to pertain to the identified objects, thereby creating a locally refined acoustic image. Utilizing this refined image, the method generates a location data set which indicates the location of the object(s) of interest. Additionally, a categorization procedure (40) may be initiated subsequently to the location identification procedure (30), the categorization procedure (40) results in a categorization data set indicating characteristics of the identified object(s).

Description

ADAPTIVE BEAMFORMING MONITORING METHOD AND SYSTEM FOR ACOUSTIC SENSOR ARRAYS Technical fieldThe present invention relates to the technical field of monitoring using acoustic arrays, specifically using adaptive beamforming for acoustic arrays and monitoring large areas. The present invention is especially suitable for the detection and categorization of objects of interest in application areas such as avalanche detection, unmanned aerial vehicle (UAV) detection, airplane collision avoidance, mobile monitoring system for deployment on vehicles, marine and city surveillance systems.Background artThe present invention pertains to the field of acoustic sensor arrays and adaptive beamforming. Adaptive beamforming is a technique in signal processing utilized for enhancing signal reception in various applications including environmental monitoring, surveillance, and object detection. The core objective of adaptive beamforming is to optimize the reception of desired signals while suppressing interference and noise, thereby improving the clarity and quality of received signals.In recent years, advancements in this field have led to the development of sophisticated beamforming techniques. For instance, US9559417B1 discusses a method involving the estimation of a spatial spectrum of waveform phenomena from sensor data, followed by determining adaptive beamforming weights using an estimated covariance matrix. This approach includes multi-taper spectral estimation and harmonic analysis to refine the beamforming process. However, while this technique enhances beamforming performance, it primarily focuses on traditional cross-correlation based beamforming relating to the initial stages of signal processing and does not encompass a comprehensive approach for real-time, dynamic environmental monitoring of multiple objects. Similarly, WO2023207047A1 describes a high-resolution acoustic spectrum estimation method using an iterative approach relying on multi-region constant beamwidth beamforming and interpolation filling. While this method marks significant progress in the field, it primarily addresses resolution enhancement rather than tracking and classification of sound sources, resulting in limitation in relation to tracking several moving objects in real time in dynamic environments. Another notable advancement, as seen in US10032464B2, involves the use of analytical broad spectrum matching using for UAV detection and classification. The matching consists of performing single channel source separation of a sample power spectral density using a non-negative matrix factorization algorithm.  SummaryIntroductionDespite aforementioned advancements, existing methods, as evidenced in patents such as US9559417B1, WO2023207047A1, and US10032464B2, exhibit certain limitations. These methods predominantly concentrate on initial signal processing stages, resolution enhancement, or specific applications like UAV detection, without adequately addressing the intricacies of real-time, dynamic environmental monitoring, especially in scenarios involving multiple moving objects. For instance, US9559417B1 enhancing beamforming performance through traditional cross-correlation based techniques but does not fully extend its capabilities to dynamic monitoring of multiple objects in real-time environments. WO2023207047A1, while marking progress in resolution enhancement, falls short in the tracking and classification of sound sources, particularly in the context of multiple moving objects. Similarly, US10032464B2, though innovative in UAV detection and classification, is limited in its broader monitoring applications and faces challenges in real-time processing due to the computational demands associated with handling multiple moving objects of interest.In large area monitoring, especially in dynamic environments, the application of adaptive beamforming is riddled with challenges. Traditional techniques, while effective under controlled conditions, struggle to address the complexities of real-time data processing, environmental variability, and simultaneous handling of multiple objects.The present invention aims to overcome the limitations of previous monitoring systems based on acoustic sensor arrays and enable for reliable and precise large area monitoring system able to handle identification of the presence, as well as the tracking and categorization – of multiple objects of interest simultaneously in real-time.First aspect: enhanced computational efficiency in object identification.The first aspect of the invention addresses the challenge of computationally efficient identification of multiple objects of interest in a dynamic environment. In a first aspect of the present invention, it is proposed a method for monitoring an area using at least one acoustic sensor array, the method comprising the steps:1. acquiring acoustic data from the at least one acoustic sensor array; and acquiring environmental data comprising at least temperature in proximity to the at least one acoustic sensor array;2. processing the acoustic data and the environmental data, and acquiring a global acoustic image, wherein said processing comprises the operation:1. adjusting the phase and amplitude of the acoustic data using a delay and sum based beamformer;3. determining, based on at least the global acoustic image if there is at least one object of interest; and if at least one object of interest has been identified, initiating a location identification procedure; wherein the location identification procedure comprises at least the following steps:1. Estimating an initial location of the at least one object of interest based on the global acoustic data and either or both of:1. a set of elevation and azimuth angles for the at least one object of interest2. a rate of change in elevation angle and azimuth angle for the at least one objects of interest; 2. Processing the acoustic data and the environmental data or their respective refinements thereof, and acquiring a locally refined acoustic image which is refined only at sections which are estimated to pertain to the at least one object of interest.3. Processing at least the locally refined acoustic image, and acquiring a location data set and / or a categorization data set; wherein the location data set is indicating a location for the at least one objects of interest; and wherein the categorization data set comprises characteristics for the at least one object of interest.The acquiring of acoustic data in step A may be done using a variety of methods known to the skilled person. The acquiring of the acoustic data typically entails converting analog signals to a digital format which is computer readable. This process involves sampling the continuous analog waveforms at discrete intervals and quantizing the amplitude of these samples to a certain bit resolution. The sampling rate and bit depth are parameters that define the temporal and amplitude resolution of the digital signal. The digital signals from all sensors in the array are aggregated and calibrated. The synchronization typically involves time-stamping the signals or using a common clock source for all ADCs. The aggregated digital signals may undergo pre-processing such as transformation from a time-domain to a frequency domain. Meaning that the acoustic data may comprise either or both of:2. a time-domain representation indicating variation of acoustic waves over time; and / or3. a frequency-domain representation indicating amplitude and phase of frequency components of acoustic waves.A time-domain representation provides advantages in relation to temporal resolution and transient analysis, which is particularly useful in application when locating exact locations of sound sources, especially in relation to short-duration sounds. However, time-domain representations does not provide as good insight into spectral content of sounds compared to frequency domain representation. Consequently, they are generally more computationally demanding when analyzing for sound signature or reducing noise. Frequency-domain representations are on the other hand highly adept at providing insight into spectral content of sounds and are thus advantageous when analyzing for sound signatures or when reducing unwanted noise. Acquiring representation of both domains typically offers a more complete understanding of the acoustic environment by capturing both temporal and spectral characteristics but does so at the cost of being more resource intense. The global acoustic image refers to a tensor consisting of spectral values for each discrete chosen combination of azimuth and elevation angle, meaning an overview of the frequency and phase content over the monitored area. The global acoustic image is designed to cover a much larger area than the locally refined acoustic image. This broad coverage necessitates a compromise on the level of detail or resolution to manage the computational load and data volume. In contrast, the local refined image focuses on a smaller, specific area where higher resolution is feasible and more critical for detailed analysis. In other words, the global acoustic image has a lower resolution than the locally refined acoustic image. As understood by the skilled person, the delay and sum based beamformer refers to a beamformer using at least a “delay and sum” technique as opposed to a cross-correlation based technique. Meaning that delay and sum based beamforming Involves calculating delays based on a physical model or a priori information about the sound source and the sensor array geometry, as opposed to cross-correlation based beamforming which involves computing the cross-correlation between signals from different sensors in the array as to evaluate the actual measured delay between sensors. A delay and sum technique first delays - meaning that for each element in the array, the signal is delayed by an amount that compensates for the difference in arrival time of the signal from a target source to that particular element; then sums - meaning that after delay alignment, the signals from all array elements are summed i.e., averaged. In other words, for each point, meaning for each discrete chosen combination of azimuth and elevation angle, the spectrum, i.e., the spectral values in a given tensor, is processed according to the following for formulaWhere is the beamformed spectrum at each point, N is the number of sensors, W are the delays and weights, and X is the spectrum for each sensor. As known to the skilled person there exist many variation of delay and sum based beamforming techniques, such as: Adaptive Beamforming Methods including, for example Minimum Variance Distortion less Response (MVDR) and the Linearly Constrained Minimum Variance (LCMV) beamformers; and Steered-Response Power (SRP) Methods including, for example Steered Response Power-Phase Transform (SRP-PHAT). By employing a delay and sum based beamformer, the invention achieves a more efficient beamforming process in relation to blind MIMO configurations. Blind MIMO in the context of acoustics refers to the propagation of an unknown amount of sources, some of which with pre-known characteristics having to be identified with a combination of multiple sensor signals. This leads to the system able to handle the identification of more objects at once, therefor increasing the precision for large area monitoring while also reducing the computational burden typically associated with full-spectrum data refinement, enabling a more advanced and precise local refinement of only section of the global acoustic image. Based on this insight, there is proposed a inventive method that in a first step refines acoustic data especially suitable for blind MIMO, such as a deconvoluted delay and sum based beamformer which has proven especially beneficial, to generate a first global acoustic image, and in a second step refine local parts of the acoustic data. A deconvoluted delay and sum based beamformer refers to a delay and sum based beamformer incorporating a subsequently applied matched deconvolving operation. Since the deconvoluted delay and sum based beamformer results in a more precise first estimation of locations of object when identifying multiple objects, the local refinement can be made to smaller areas of the global image. This decreases required computational power and allows for more advanced analysis and in the monitoring system being able to handle more objects of interest and thus cover a greater area for monitoring.Embodiments for Deconvolution of global acoustic image.In the domain of acoustic monitoring, particularly in applications such as avalanche detection, UAV surveillance, and city-wide noise mapping, traditional signal processing methods are based on well-established physical models that have been developed and refined over decades. Traditional techniques in this sense refer to analytic techniques not relying on neural networks, such as analytical atmospheric absorption models for correction in relation to atmospheric damping, geometric ray tracing for correction in relation to reflection analysis for correction in relation to ground reflection, doppler shift compensation algorithms for correction in relation to doppler effect, spatial filtering and signal phase correction for correction in relation to Lloyd mirror effect and spectral subtraction and noise cancellation for correction in relation to acoustic masking. The accuracy and reliability of the acoustic signal analysis in monitoring application is critical and due to the traditional methods relying on physical principles offer more transparent operations i.e., decision-making, they are generally perceived as offering a high degree of certainty and repeatability. This perception is further endorsed by the fact that the outcomes of these methods can be anticipated and validated against known acoustic behavior.In one proposed embodiment, the processing in step B further comprises the operation:1. deconvolving the acoustic data with respect to a first set of factors influencing acoustic propagation characteristics using a neural network, preferably a recursive neural network. As understood by a skilled person, what factors are compensated or corrected for in deconvolution is intricately connected to the operational characteristics of the antecedently performed beamforming. In other words, the factors for optimization or correction in the deconvolution process are selected based on the refinements made by the beamforming technique. Due to the iterative nature of the proposed method, it is preferable for the deconvolving operation in step B to be optimized for spatial invariance rather than frequency invariance. level Fluctuations induced by spatial variance levels can lead to unreliable detection, as weaker signals might be missed or interpreted as noise, while stronger signals could be falsely identified as different sources. Optimizing for spatial invariance minimizes this risk since maintaining consistent levels helps in achieving a better SNR. A spatial consistent level, irregardless of direction, ensures that the signal stands out against the background noise, making it easier to identify and locate the target. By focusing on spatial invariance, the deconvolution process provides a stable and reliable foundation for subsequent signal analysis and interpretation.In one embodiment, the deconvolution process in step B addresses a first set of factors including at least one of: minimized spatial smearing due to beampatterns, increase in SNR and / or reduced frequency coloring. In a preferred embodiment, special emphasis is placed on reducing the spatial smearing of the conventional beamformer, meaning that in this preferred embodiment the first set of factors include at least Signal to Noise ratio. Optimizing the deconvolution process in step B to address the spatial smearing effectively reduces the influence of ambient noise and increases the clarity and discernibility of the acoustic sources.In one alternative embodiment, the deconvolution process addresses a set of factors including at least White-Noise-Gain. The proposed monitoring method incorporates a neural network (NN) to significantly boost computational efficiency, especially in acoustic-based monitoring systems handling multiple objects. Traditional methods have limitations in dealing with non-linear distortions, varying environments, and simultaneous tracking of multiple objects. The proposed invention entailing; refining a global data set coarsely and subsequently selectively refining local parts of the acoustic data, achieves synergistic effects when combined with NNs for deconvolving acoustic data in large area monitoring.The proposed embodiment is based on the insight is that NNs, while excelling in associating and analyzing temporal or spatial patterns, may not be as effective in separating signal characteristics like frequency and spatial attributes as more traditional methods. Therefore, the present invention suggest using NNs to deconvolve a initial overview of the frequency content over the monitored area, i.e. the global acoustic image. The global acoustic image is used to identify the presence of objects of interest and to make a coarse estimation of their location. Consequently, the NN can be tailored towards this specific purpose of identifying the presence of potential objects of interest and allocate the tracking and / or categorization of the object to be computed in a manner suitable for those purposes. This method allows for a computationally efficient process, leading to more precise determination of location as well as categorization through local refinement of the acoustic image.One key advantage of this approach is the synergistic effect of the combination of NNs with the coarse-to-fine refinement strategy. This synergy is dynamically enhanced by the adaptive learning capabilities of NNs. For instance, the neural network may input the locally refined acoustic image and compare it with the corresponding area in the global acoustic image and use this comparison to improve its performance. This results in a system that not only operates efficiently but also increases its accuracy and effectiveness over time, particularly in complex and large-scale acoustic monitoring scenarios.For acoustic based monitoring systems to maintain precision when handling multiple objects computational efficiency is particularly vital in the steps leading up to the identification of presence of objects of interest within the monitored area. Generating the initial global acoustic image is a critical step for the identification of presence of objects of interest. NNs contribute to increasing the efficiency of this step by adeptly handling data at lower resolutions. They employ broad pattern recognition capabilities, which allows them to infer or "fill in the blanks" in the data, drawing on their accumulated learnings and insights. This ability of NNs to interpret and extrapolate from incomplete or lower-resolution data not only conserves computational resources but also ensures that the system remains precise and effective in its monitoring and tracking tasks. Consequently, the use of NNs in these systems enhances their ability to process large volumes of data while maintaining high accuracy in object detection and tracking.Consequently, employing a neural network in the processing of step B leads to a more efficient and effective monitoring system capable of handling a larger number of objects and covering a wider area with reduced computational load by adeptly handling non-linear distortions and adapting to variable environmental conditions.Embodiments accounting for factors influencing acoustic propagation characteristics.In a large area, the acoustic waves from a sound source will travel through a variety of different environments, each with its own unique profile. This can cause the waves to refract, reflect, and scatter, which can distort the received signal and make it more difficult to detect and localize the sound source.However, in a smaller area, the acoustic waves from a sound source will travel through a more homogeneous environment, and the effects of factors such as temperature will be less pronounced. As a result, it may not be necessary to take changes in temperature into account enabling correction factors to be based on a simplified model of the environment and be fixed during runtime.While there exist many known techniques to take into consideration acoustic propagation characteristics, it is a technical problem to realize that there is a technical advantage to doing so. Consequently, it is not readily obvious, and at times not even desired - to consider factors influencing acoustic propagation characteristics in monitoring systems relying on acoustic array sensor. In one proposed embodiment the environmental data may comprise information that can be used to determine acoustic propagation characteristics additional to that of temperature. What the environmental data comprises is dependent on the use case and what sensors are used. When monitoring acoustics, i.e. pressure wave variations, in gas, e.g. air, such as for the use case of UAV monitoring, the environmental data may include, for example, wind speed, wind direction, and / or humidity. When monitoring acoustic in solids and for use cases such as avalanche detection, the environmental data may include information relating to density, depth, wave type, snow density and structure, moisture content. When monitoring acoustics in fluid, such as for underwater monitoring, the environmental data may include salinity, depth (i.e. pressure), water currents, thermoclines, turbulence and other water movement. Temperature is given special emphasis due to its notable influence on acoustic propagation characteristics. It is preferably taken into account by adjusting the set of weighting factors and time delays in the delay and sum beamformer. The acquisition of information influencing acoustic propagation characteristics, such as temperature data, may be achieved through positioning ambient sensors, such as a temperature sensor - in proximity to the acoustic sensor array, or through communication with an API to retrieve area-specific data. Alternatively, factors influencing acoustic propagation characteristics may be derived from the acoustic data gathered. In other words, environmental data may be acquired from at least one ambient sensor, a secondary interface and / or it may be derived from acoustic data. Due to computational requirements and precision, it is preferable to acquire environmental data from ambient sensor and / or a secondary interface.By taking factors influencing acoustic propagation characteristics into account, the weight factors and correction factors can be adjusted to compensate for these distortions and improve the accuracy of the monitoring system. This is particularly important for applications such as UAV detection, where it is critical to accurately track the movement of UAV’s over large distances.Temperature is the most influential environmental factor influencing acoustic propagation characteristics in air. It is also relatively cheap and easy to monitor. By accounting for temperature fluctuations, the proposed embodiment dynamically adjusts the weight factors and correction factors employed in beamforming algorithms. This adaptation effectively compensates for temperature-induced distortions, resulting in improved sound source detection and localization capabilities.Based on this insight, there is proposed an inventive method that in a first step refines acoustic data especially suitable for blind MIMO, such as a delay and sum based beamformer, to generate a global acoustic image, and in a second step refine local parts of the acoustic data. Since environmental data is taken into account, the first estimation of where the multiple objects are is more precise, leading to a more precise local refinement. This decreases required computational power, allows for more advanced analysis and results in the monitoring system being able to handle more objects of interest and thus cover a greater area for monitoring.Embodiments for detecting presence of potential objects of interest.In acoustic monitoring systems, such as those used for avalanche detection or urban noise mapping, accurately detecting the presence of objects of interest within a global acoustic image is paramount. Traditionally, analytical techniques for presence detection, such as Spectral Kurtosis analysis, Mel-frequency cepstral coefficients (MFCCs), or analytical broad-spectrum matching, have been widely used. MFCCs are generally chosen for their ability to handle complex noise environments and effective in categorizing sounds. While Spectral Kurtosis may be used when detecting non-Gaussian and transient signals is more important. However, these methods may not always provide the desired level of accuracy and computational efficiency, especially in dynamic and complex environments where multiple objects of interest may be present simultaneously. Furthermore, such traditional methods may struggle to adapt to varying environmental conditions and can be computationally intensive.The proposed invention involves, in step C, determining if there is at least one object of interest based on at least the global acoustic image. As known to the skilled person there exist a variety of techniques to determine the presence of objects of interest. As also known, the acoustic data may be complemented by, for example, video data or RADAR data. The determining of presence in step C may utilize traditional techniques such as kurtosis analysis, Mel-frequency cepstral coefficients (MFCCs), or analytical broad-spectrum matching. Alternatively, a neural network-based approach may be employed such as: feedforward networks, or a subvariant thereof, for example Multi-Layer Perceptrons (MLP) or Convolutional Neural Networks (CNNs); Neural Networks (NNs) or a subvariant thereof; Recurrent Neural Network (NN), or a subvariant thereof, for example Long Short-Term Memory (LSTM) or Gated Recurrent Unit (GRU); or a combination of these networks. In some cases it is preferable to use a hybrid approach and employ both traditional techniques and neural networks.In one preferred embodiment step C comprise determining if there is at least one object of interest based at least on the global acoustic image using a neural network. The choice of a specific neural network depends on the design requirements of the system. For instance, RNNs are highly capable of handling temporal information, which is beneficial for modeling the temporal dynamics of sound associated with fast-moving objects like UAVs. CNNs, in contrast, are adept at handling spatial information, potentially enabling more advanced feature extraction from the spatial representation of the sound. Within this framework, neural networks can be integrated in various ways: sequentially, in parallel, or using a hybrid integration such as a cascade or recurrent structure. The proposed embodiment allows for a tailored system design based on specific requirements, which is especially advantageous in large area monitoring applications. Depending on the monitoring needs, the system can for example be adjusted to focus on temporal dynamics and / or detailed spatial analysis for specific operations.Furthermore, the possibility of integrating NNs such as CNNs and / or RNN in isolation, sequentially, in parallel, or in a hybrid structure offers immense flexibility. This aligns with the objectives of previous aspects and embodiments by allowing for a wide range of monitoring scenarios and improving overall system performance.Embodiments for local refinementStep C, 2 in the proposed method involves processing the acoustic data and the environmental data or their respective refinements thereof and acquire a locally refined acoustic image which is refined only at sections which are estimated to pertain to the at least one object of interest.The locally refined acoustic image refers to an enhanced, detailed subset of the global acoustic image. Meaning that, unlike the global acoustic image, which covers a larger area with a compromise on detail or resolution, the locally refined acoustic image focuses on a smaller, specific area or sections where a higher resolution is both feasible and more critical for detailed analysis. The locally refined acoustic image is preferably characterized by higher resolution and focused processing. Higher resolution refers to a more granular and detailed acoustic image. Focused processing preferably refers to at least one of:2. Enhanced spectral and temporal resolution; meaning that the resulting image contains enriched spectral and temporal data for targeted sections. This means that for each discrete chosen combination of azimuth and elevation angle within these sections, the spectral values and / or time data are processed to reveal more detailed information, like peak frequency shifting and short time level variations, and temporal changes that might be crucial for accurately identifying, tracking, or categorizing the objects of interest.3. Spatially and Temporally Targeted; meaning that the refinement may be spatially targeted to specific sections, which are determined based on the initial analysis of the global acoustic image. It can also be temporally focused, emphasizing changes over time within these sections, which is particularly important for moving or dynamically changing objects.4. Customized for Object of Interest: meaning that the processing and refinement in the locally refined acoustic image are customized based on the characteristics of the object of interest. For instance, if the object of interest is known to emit sound at specific frequencies, the refinement might focus on enhancing those frequencies within the chosen sections.The processing in step C, 2. refers to operations applied the acoustic data, at least some of which are based on the environmental data. These operations may include, for example, digital filtering, signal enhancement, noise reduction, and data transformation techniques. The processing in step C, 2 aims to improve the quality and intelligibility of the data, making it more conducive for further analysis.The term "a refined combination thereof" refers to the integration of processed acoustic and environmental data to create a composite dataset. This refinement may involve aligning and synchronizing data points from both datasets, which may be followed by applying techniques such as data fusion and cross-correlation to enhance the overall quality and relevance of the information. The refinement process in these estimated sections involves applying targeted signal processing techniques, such as adaptive filtering or beamforming algorithms, tailored to the specific use case and characteristics of the object of interest. The objective is to enhance the signal quality and extract detailed information from these sections without distorting the overall acoustic image. A variety of alternative methods are available for the processing in Step C, 2, i.e., the refinement process for acquiring the locally refined acoustic image. As understood by the skilled person, these alternative methods offer various advantages in different monitoring scenarios and can be selected based on the specific needs of the application.Examples of alternative methods include: Adaptive Filtering Techniques which dynamically adjust filtering parameters in response to changing acoustic conditions, beneficial for non-stationary noise conditions. Time-Frequency Analysis, using methods like Short Time Fourier Transformation (STFT) or wavelet transforms, beneficial for in-depth temporal and spectral analysis. Machine Learning-Based Classification, such as CNNs or SVMs, effectively processes distinct acoustic signatures of objects. Directional Beamforming which entails utilizing spatial information for signal enhancement in known sound source locations. Lastly, Spatio-Temporal Filtering merges spatial and temporal data analysis, offering significant benefits in dynamic settings, albeit with greater complexity. Each method, as understood by the skilled person, offers unique advantages and is chosen based on the specific demands of the monitoring scenario.Embodiments for localizationThe locally refined acoustic image may be processed as to determine a precise location of the object of interest and / or it could be processed as to determine characteristics of the object of interest, i.e., categorizing it. What kind of refinement process is used is preferably adapted to what information about the object of interest is desired. In traditional acoustic based monitoring system it is typical to implement one refinement process using one of the above mentioned alternative methods for refinement, and to derive insights into both location and category from the resulting refined image. However, due to the modular approach proposed according to the present invention, it is possible to employ refinement processes tailored towards the extracting the desired information.Based on this insight, it is proposed a embodiment which acquires location data set in step C, 3 based on a locally refined acoustic image specifically adapted to be used for localization of sound sources. In this embodiment, the processing in step C, 2. comprises performing at least one of the following operations on the acoustic data only at sections which are estimated to pertain to the at least one object of interest:1. adjusting the phase and amplitude of the acoustic data or a refinement thereof using a first set of weights and time delays; and / or2. deconvolving the acoustic data or a refinement thereof with respect to a second set of factors; As understood by a skilled person, what the second set of factors are designed to compensate or corrected for in the deconvolution is intricately connected to the operational characteristics of the antecedently performed beamforming, e.g., the first set of weights and time delays. In other words, the first set of factors which are compensated or corrected for in the deconvolution process are selected based on the refinements made in the beamforming. The locally refined image is in this proposed embodiment adapted for the specific purpose of localization and tracking of objects. In one proposed embodiment the deconvolving is optimized for spatial invariance rather than frequency invariance. Level fluctuations induced by spatial variance can lead to unreliable localization or tracking as weaker signals, or signals that change their relative distance or angle to the sensor might be missed or interpreted as noise, while stronger signals could be falsely identified as different sources. Optimizing for spatial invariance minimizes this risk since maintaining consistent levels helps in achieving a better SNR. A spatial consistent level, irregardless of direction, ensures that the signal stands out against the background noise, making it easier to identify and locate the target. By focusing on spatial invariance, the deconvolution process provides a stable and reliable foundation for subsequent signal analysis and interpretation.In one proposed embodiment, the second set of factors preferably relate to static source propagation effects and preferably comprises at least one of: Lioyd mirror effect, phantom sources and / or ground reflection.Embodiments for categorizationThe processing in step C, 3 can be performed as to either acquire a location data set or a categorization data set, or both. This flexibility allows for the proposed method to be adapted to a variety of application areas. For example, when tracking avalanches or wildlife, the estimated location might not need to be very precise but deriving characteristics about the object of interest is paramount. Therefore, in some embodiments, it may be more suitable to not perform the local refinement for the purpose of localization but instead derive the estimated location from the global acoustic data as to allocate further computing resources to the analyses of the object of interest in relation to its characteristics. In other cases, such as when monitoring for very transient events like gunshots, the characteristics of the gunshot might not be as important as determining only the exact location of the sound. In such cases it may be suitable to perform the local refinement tailored to the purpose of sound localization or tracking. When both location and characteristics of the object are of importance, such as when monitoring for UAV detection or airplane collision prevention, the locally refined acoustic image acquired in step C, 3 may be tailored towards deriving location of sound sources, and a subsequent refinement may be made that is tailored towards categorization. However, even in such configurations the alternative to derive an initial categorization data set in step C, 3 could be advantageous. For example, if a categorization procedure is imitated subsequent to the location identification procedure, the initial categorization data set may comprise information advantageous to configure a second set of weighting factors and / or deconvolution in regards to a third set of factors used in a local refinement performed in the subsequent categorization procedure.In a preferred embodiment only a location data set is acquired in step C, 3, and a subsequent step acquires the categorization data set. It is proposed that in this embodiment, a categorization procedure is initiated after at least one iteration of the location identification procedure is completed. The categorization procedure comprises step D, entailing processing the acoustic data or a refinement thereof at sections which are estimated to pertain to the at least one object of interest and acquiring a categorization data set indicating a category type for the at least one objects of interest. The categorization data set could comprise any information relating to an object of interest which is relevant for the specific use case. For example, the information may include:5. an acoustic signature profile which can include parameters such as frequency spectrum characteristics, intensity patterns, temporal variations, and specific acoustic fingerprints that are unique to certain types of objects;6. spectral features such as Mel-frequency cepstral coefficients (MFCCs), spectral centroid, spectral bandwidth, spectral flatness, and / or spectral roll-off;7. temporal dynamics, meaning information indicating temporal evolution of the acoustic signal, which might include attack time, decay, sustain, release patterns, and temporal variations in intensity and frequency;8. Spatial characteristics representing the spatial distribution of the sound source, such as directionality, apparent size based on acoustic dispersion, and movement patterns based on changes in spatial attributes over time;9. Harmonic and percussive elements for identification of harmonic (tonal) and percussive (noisy / transient) components within the sound;10. Behavioral patterns, meaning information related to patterns of movement, speed, acceleration, and trajectory, which can be used to infer the nature of the object (e.g., vehicle, animal, human);11. Cross-referenced data, meaning information that correlates acoustic data with other types of sensor data, such as visual (from cameras), thermal, or radar, to enrich the categorization process.To what extent the above examples are included in the categorization data set depends on the specific use case. For example, for UAV detection the categorization data set may be designed to incorporate an acoustic signature profile, encompassing parameters like frequency spectrum characteristics, intensity patterns, and specific acoustic fingerprints unique to different UAV models. This profiling can include an analysis of spectral features such as Mel-frequency cepstral coefficients (MFCCs), spectral centroid, and spectral roll-off, particularly relevant to identifying rotor noise and frequency variations attributed to UAV size and type. Behavioral patterns indicating UAV activity, such as hovering, ascending, descending, or rapid lateral movements, can be inferred from temporal dynamics like attack time, decay, and variations in intensity and frequency. Operational characteristics may be estimated, including UAV size, weight, and possible payload, through analysis of harmonic and percussive elements within the UAV's sound. Anomaly detection in acoustic patterns can also be integrated, detecting unexpected changes in UAV sounds that may indicate abnormal operation or potential threats.In avalanche detection, the categorization data set can include avalanche acoustic indicators, identifying subtle acoustic signs like cracking sounds or small-scale snow movements through temporal dynamics and spectral features. This data set can also characterize the progression of an avalanche using information on temporal evolution, such as speed and direction of movement, based on changing acoustic patterns. Post-event, the processing may entail analyzing sounds to assess stability and the possibility of subsequent avalanches. Additionally, environmental impact assessment can be performed, estimating the avalanche's size and impact area based on acoustic intensity and spread. This can involve examining spatial characteristics and temporal variations in the acoustic signals to understand the event's dynamics.For city surveillance, the categorization data set may be tailored to include urban sound categorization, which might include classifying sounds into categories such as traffic, human voices, alarms, construction, and emergency services. To achieve this the processing step might entail analyzing temporal sound patterns, examining variations in sounds like traffic density or crowd noise over different times of the day or week. Event detection and characterization can identify specific events like car accidents or public gatherings based on unique sound signatures. This process may leverage spatial characteristics and behavioral patterns to infer the nature of various sound sources within the city. It is preferable that the processing in step D comprises performing at least one of the following operations on the acoustic data only at sections which are estimated to pertain to the at least one object of interest:1. adjusting the phase and amplitude of the acoustic data or a refinement thereof using a second set of weights and time delays; and / or2. deconvolving the acoustic data or a refinement thereof with respect to a third set of factors.As understood by a skilled person, what the third set of factors are designed to compensate or corrected for in the deconvolution is intricately connected to the operational characteristics of the antecedently performed beamforming, e.g., the second set of weights and time delays. In other words, the third set of factors which are compensated or corrected for in the deconvolution process are selected based on the refinements made in the beamforming. The locally refined image is in this proposed embodiment adapted for the specific purpose of localization and tracking of objects. In one proposed embodiment the deconvolving is optimized for spatial invariance rather than frequency invariance.In one preferred embodiment the third set of factors relate to dynamic source propagation effects and comprises preferably doppler effect and / or atmospheric damping.Embodiments for applications areasIn the context of the present application, acoustic data refers to the information captured from sound waves traveling through any medium, such as fluids (e.g. water), solids (e.g. earth), and gas (e.g. air). The sound waves, i.e., the oscillation in pressure, particle displacement, particle velocity and / or density, may sensed in a number of different ways depending on the given medium. What specific type of sensor is suitable depends largely on the given medium but could also be adapted for specific use cases. For example, when monitoring an area underwater a hydrophone array is suitable. If monitoring for objects moving in the air or on a surface on land, a microphone arrays is suitable. In avalanche or earthquake monitoring, i.e., when capturing sound waves moving through earth, a seismic sensor array implemented with geophone may be preferable. A seismic sensor might refer to any sensor capable of detecting movement within solids such as a load cell, laser doppler vibrometer and / or accelerometer array.In a second aspect of the present invention it is proposed a method for aircraft surveillance incorporating the steps of the first aspect or an embodiment thereof. In this aspect the acoustic data is preferably acquired from at least one microphone array. In one preferred embodiment of the second aspect the method is specifically adapted for UAV surveillance, in other words drone detection. In other embodiments the method may survey for any aircraft which for the purpose of preventing collisions. In a third aspect of the present invention it is proposed a method for geophysical mass flow monitoring incorporating the steps of the first aspect or an embodiment thereof. In this aspect the acoustic data is preferably acquired from at least one seismic sensor array. The geophysical mass flows monitored for may be objects such as avalanches, landslides or soil liquefaction. In a fourth aspect of the present invention it is proposed a method for underwater monitoring incorporating the steps of the first aspect or an embodiment thereof. In this aspect the acoustic data is preferably acquired from at least one hydrophone array. In a fifth aspect of the present invention it is proposed a method for land surface surveillance incorporating the steps of the first aspect or an embodiment thereof. In this aspect the acoustic data is preferably acquired from at least one microphone array. Land surface surveillance refers to monitoring for sound sources located on a surface of land, this may include application such as: studded tire monitoring, shooting / sniper detection, intrusion surveillance. In a sixth aspect of the present invention, there is provided a method performed by at least one processing circuit for monitoring an area using at least one acoustic sensor array, the method comprising the steps:1. acquiring acoustic data from the at least one acoustic sensor array; and acquiring environmental data comprising at least temperature in proximity to the at least one acoustic sensor array, the environmental data being indicative of how the environment in the monitored area influence acoustic propagation characteristics;2. processing the acoustic data and the environmental data to acquire a global acoustic image, wherein said processing comprises the operation:4. adjusting the phase and amplitude of the acoustic data using a delay and sum based beamformer;3. determining, based on at least the global acoustic image if there is at least one object of interest; and if at least one object of interest has been identified, initiating a location identification procedure; wherein the location identification procedure comprises at least the following steps:estimating an initial location of the at least one object of interest based on the global acoustic image and either or both of:1. a set of elevation and azimuth angles for the at least one object of interest derived from the global acoustic image; and,2. a rate of change in elevation angle and azimuth angle for the at least one objects of interest derived from the global acoustic image; processing the acoustic data and the environmental data or processing the global acoustic image using a neural network to acquire a (e.g., at least one) locally refined acoustic image which is refined only at sections which are estimated to pertain to the at least one object of interest;processing at least the (e.g., at least one) locally refined acoustic image, and acquiring a location data set and / or a categorization data set; wherein the location data set is indicating a location for the at least one objects of interest; and wherein the categorization data set comprises characteristics for the at least one object of interest.In a seventh aspect of the present invention, there is provided a system for monitoring an area using at least one acoustic sensor array, wherein the system comprises:1. an information gathering module adapted to:1. receive acoustic data from the at least one acoustic sensor array; 2. receive environmental data from at least one ambient sensor and / or a secondary interface; wherein said environmental data comprises at least temperature in proximity to the at least one acoustic sensor array, said environmental data being indicative of how the environment in the monitored area influence acoustic propagation characteristics; 2. a global acoustic image module adapted to generate a global acoustic image based on processing the acoustic data and the environmental data, wherein the said processing to generate the global acoustic image comprises:3. acquiring a phased array output by using a delay and sum based beamformer module; wherein the delay and sum based beamformer module is adapted to adjust the phase and amplitude of the acoustic data;4. acquiring the global acoustic image by deconvolving the phased array output using a first deconvolution module;3. an identification module adapted to analyze the global acoustic image to identify the presence of potential objects of interest, wherein the identification module initiates a location identification procedure if at least one object of interest is identified;4. a spatial orientation module adapted receive the global acoustic image and generate based thereon: a first spatial orientation data set and a second spatial orientation data; wherein said first spatial orientation data set comprises current elevation and azimuth angles for the at least one objects of interest; wherein said second spatial orientation data set comprises current apparent heading and angular speed for the at least one objects of interest, the second spatial orientation data set being based on a rate of change of elevation angle and / or azimuth angle for the at least one objects of interest;5. a local refinement module adapted to process the acoustic data and the environmental data or process the global acoustic image using a neural network to generate a locally refined acoustic image which is refined only at sections which pertain to at least one object of interest; said sections being determined based on the first spatial orientation data set and / or the second spatial orientation data set;6. a location module set adapted to generate a location data set and / or a categorization data set based on the locally refined acoustic image wherein the location data set is indicating a location for the at least one objects of interest, and wherein the categorization data set comprises characteristics for the at least one object of interest. Brief description of drawingsFigure 1 to 5 illustrates steps according to the proposed aspects and embodiments of the present invention. To facilitate understanding, optional features are illustrated in dashed lines.Figure 1 illustrates step A and the acquiring of acoustic data and environmental data according to one embodiment of the present invention.Figure 2 illustrates step B and the processing performed to acquire the global acoustic image according to one embodiment of the present invention.Figure 3 illustrates the location identification procedure according to one embodiment of the present invention.Figure 4 illustrates the categorization procedure according to one embodiment of the present invention.Figure 5 illustrates an overview of the sequence in which steps may be performed according to one embodiment of the present invention.Figure 6 and 7 illustrates a system according to one embodiment of the present invention. Figure 6 shows an information gathering module, a global acoustic image module, an identification module and arrows illustrating the communication therebetween.Figure 7 shows an information gathering module, a spatial orientation module, a local refinement module, a location module, a signature refinement module, a categorization module and arrows illustrating the communication therebetween.Detailed descriptionFigure 1 illustrates an embodiment of the method according to the present invention showing step A entailing acquiring 10 at least acoustic and environmental data. In the illustrated embodiment the step is subdivided into two key operations. The first is acquiring acoustic data 11 from at least one acoustic sensor array. This operation typically involves the conversion of analog signals into a digital format that computers can read, a process that may include sampling of continuous analog waveforms at discrete intervals and quantizing their amplitudes to a certain bit resolution. The digital signals may then be aggregated, synchronized, and pre-processed to provide a time-domain representation of acoustic waves over time and / or a frequency-domain representation indicating the amplitude and phase of frequency components.The second operation within Step A, acquiring environmental data 12 encompasses collecting information indicating how the environment in the monitored area influence acoustic propagation characteristics. Preferably the environmental data encompasses at least temperature. Depending on the specific use case and the types of sensors employed, environmental data may also include a variety of other measurements, such as wind speed and direction, humidity, snow density, moisture content, salinity, water currents, and more. The collection of this data can be accomplished through ambient sensors placed near the sensor arrays or via a secondary interface like an API. Additionally, environmental data can sometimes be inferred directly from the acoustic data itself.Now turning to Figure 2, showing one proposed embodiment of step B, meaning the processing 20 performed as to acquire the global acoustic image. The global acoustic image refers to a tensor consisting of spectral values for each discrete chosen combination of azimuth and elevation angle, meaning an overview of the frequency and phase content over the monitored area. In the illustrated embodiment the step is delineated into three distinct operations.The initial operation, adjust phase & amplitude of acoustic data using delay and sum based beamformer 21, is an operation where each element in the sensor array delays the received signal to compensate for the variance in signal arrival time from each discrete chosen combination of azimuth. Following the alignment of these delays, the signals are then coherently summed across all elements, effectively averaging them to enhance the target signal while reducing noise and other artifacts.Beneath this, dashed lines enclose an optional step: deconvolving the acoustic data or a refinement thereof, with respect to a first set of factors 22. The specific operations performed in the deconvolution in step B is dependent on the prior beamforming step 21. In a preferred embodiment the deconvolution involves the use of a neural network. This operation aims to correct for factors affecting acoustic propagation.The final operation in step B is to acquire the global acoustic image 23, in other words a spectral overview is obtained over the monitored area. Preferably, this global acoustic image is a representation with lower resolution than a locally refined image due to the trade-offs made to handle large data volumes and computational demands effectively.Now turning to Figure 3 illustrating one embodiment of the location identification procedure 30.The illustrated location identification procedure 30 starts with estimating an initial location 31. The initial location estimation is a preliminary determination of the object's position, it could be a direction in reference to an acoustic sensor array or a triangulated location derived from multiple acoustic sensor arrays. This initial estimated location is derived from the global acoustic image previously acquired.Subsequently, the location identification procedure 30 is configured to acquire a locally refined acoustic image 32. This a more nuanced and detailed acoustic image, focusing on a smaller, specific area thought to contain the objects of interest. As opposed to the global acoustic image the locally refined acoustic image involves higher resolution and focused processing to discern finer details within the target sections.The locally refined acoustic image may be acquired 32 by performing the two sub-operations, illustrated in dashed lines for emphasizing that these are optional. The first, Adjusting the phase and amplitude of the acoustic data or a refinement thereof using a first set of weights and time delays 321, refers to the fine-tuning of the received acoustic data to enhance clarity and accuracy of the object's acoustic signature. The second, Deconvolving the acoustic data or a refinement thereof with respect to a second set of factors 322, involves a correction process to reverse effects of acoustic propagation phenomena that may have distorted the original signal.The illustrated location identification procedure involves acquiring location data set 33 based on the locally refined acoustic image. Additionally, the procedure may involve acquiring an initial categorization data set 34. Now turning to Figure 4 showing one proposed embodiment for the categorization procedure 40. In the illustrated embodiment, the categorization procedure 40 commences with by acquiring a locally refined acoustic image 42. Preferably the locally refined acoustic image 42 is a different from the locally refined acoustic image 32 acquired in the location identification procedure 30. This is preferably as it enables for the local refinement to be tailored towards localization respectively categorization. The locally refined acoustic image 41 acquired in the categorization procedure focuses on the specific sections of the acoustic data pertinent to the object of interest.Nested within this step of acquiring the locally refined acoustic image 41 two optional sub-operations, denoted by dashed outlines, that delve into the details of the refinement process. Adjusting the phase and amplitude of the acoustic data or a refinement thereof using a second set of weights and time delays 411 fine-tunes the data by aligning the time and intensity of the received acoustic signals, ensuring that the features of the object are clearly presented and not obscured by background noise or other distortions.The subsequent sub-operation, deconvolving the acoustic data or a refinement thereof with respect to a third set of factors 412, represents a correction procedure aimed at counteracting the effects that interfere with the propagation of the acoustic signals, such as Doppler shifts or atmospheric absorption. In the illustrated embodiment the categorization procedure 40 ends with the step of acquiring a categorization data set 43 which may include information classifying the object based on its unique acoustic signature and other relevant attributes such as the object's behavior, movement, and any anomalies present in its acoustic profile.Now turning to Figure 5 illustrating one proposed configuration of the sequence of operations according to the present invention. The illustration shows an overview encapsulating several steps that collectively contribute to the identification and characterization of objects within the monitored area.Step A entails acquiring 10 acoustic and environmental data. Following this the acquired data is processed 20 in step B, this step involves the manipulation of data to produce a global acoustic image. Subsequent to step B, it is determined 25 whether or not any identified sound sources in the global acoustic image are objects of interest. The determination of the presence of objects of interest can incorporate various techniques, from traditional signal processing methods to advanced neural network-based analysis.If the outcome of this determination is affirmative, meaning that objects of interest have been identified, the flow proceeds to the location identification procedure 30 which is adapted to acquire at least a location data set 33 based on a subsequently acquired locally refined acoustic image 32. In figure 5 there is shown an embodiment in which the categorization procedure is initiated after the location identification procedure, however, in other embodiments, the categorization producer may be initiated as to be performed in parallel with the location identification procedure. Step A 10, step B 20 and determining if there are objects of interest in the area is preferably performed continuously, while the location identification procedure 30 may be initiated first when a object of interest having been identified. The categorization procedure 40 may be initiated subsequent to the location identification procedure or based on certain requirements, such as requirements pertaining to type of object of interest which may be derived from an initial categorization data set 34 acquired in the location identification procedure.Figure 6 and 7 illustrates a system for monitoring according to one embodiment of the present invention. The illustrated embodiment the system comprises: an information gathering module 100, a global acoustic image module 220, a identification module 300, a spatial orientation module 400, a local refinement module 500, a location module 600, signature refinement module 700 and a categorization module 800. The system according to the present invention may be configured in a number of different ways. Turning now to figure 6 illustrating the communication between the information gathering module 100, the global acoustic image module 200 and the identification module 300.The information gathering module 100 is adapted to:1. receive acoustic data 110 from the at least one acoustic sensor array; and,2. receive environmental data 120 from at least one ambient sensor and / or a secondary interface; wherein said environmental data 120 comprises at least temperature in proximity to the at least one acoustic sensor array.In one proposed embodiment the environmental data 120 may comprise information that can be used to determine acoustic propagation characteristics additional to that of temperature. What the environmental data 120 comprises is dependent on the use case and what sensors are used. When monitoring acoustics, i.e. pressure wave variations, in gas, e.g. air, such as for the use case of UAV monitoring, the environmental data 120 may include, for example, wind speed, wind direction, and / or humidity. When monitoring acoustic in solids and for use cases such as avalanche detection, the environmental data 120 may include information relating to density, depth, wave type, snow density and structure, moisture content. When monitoring acoustics in fluid, such as for underwater monitoring, the environmental data 120 may include salinity, depth (i.e. pressure), water currents, thermoclines, turbulence and other water movement. Temperature is given special emphasis due to its notable influence on acoustic propagation characteristics. It is preferably taken into account by adjusting the set of weighting factors and time delays in the delay and sum beamformer module 210. The acquisition of information influencing acoustic propagation characteristics, such as temperature data, may be achieved through positioning ambient sensors, such as a temperature sensor - in proximity to the acoustic sensor array, or through communication with an API to retrieve area-specific data. Alternatively, factors influencing acoustic propagation characteristics may be derived from the acoustic data 110 gathered. In other words, environmental data 120 may be acquired from at least one ambient sensor, a secondary interface and / or it may be derived from acoustic data. Due to computational requirements and precision, it is preferable to acquire environmental data 120 from ambient sensor and / or a secondary interface.3. The global acoustic image module 200 is adapted to generate a global acoustic image based on the acoustic data 110 and the environmental data 120. In the illustrated embodiment the global acoustic image module 200 performs the following:1. acquiring a phased array output by using a delay and sum based beamformer module 210; wherein the delay and sum based beamformer module 210 is adapted to adjust the phase and amplitude of the acoustic data 110; and,2. acquiring the global acoustic image by deconvolving the phased array output using a first deconvolution module 220.The global acoustic image refers to a tensor consisting of spectral values for each discrete chosen combination of azimuth and elevation angle, meaning an overview of the frequency and phase content over the monitored area. The global acoustic image is designed to cover a much larger area than the locally refined acoustic image. This broad coverage necessitates a compromise on the level of detail or resolution to manage the computational load and data volume. In contrast, the local refined image focuses on a smaller, specific area where higher resolution is feasible and more critical for detailed analysis. In other words, the global acoustic image has a lower resolution than the locally refined acoustic image. As understood by the skilled person, the delay and sum based beamformer module 510 refers to a beamformer using at least a “delay and sum” technique as opposed to a cross-correlation based technique. Meaning that delay and sum based beamforming 510 involves calculating delays based on a physical model or a priori information about the sound source and the sensor array geometry, as opposed to cross-correlation based beamforming which involves computing the cross-correlation between signals from different sensors in the array as to evaluate the actual measured delay between sensors. As understood by a skilled person, what factors are compensated or corrected for in deconvolution is intricately connected to the operational characteristics of the antecedently performed beamforming. In other words, the factors for optimization or correction in the deconvolution process are selected based on the refinements made by the beamforming technique. Due to the iterative nature of the proposed method, it is preferable for the first deconvolving module 220 is adapted to be optimized for spatial invariance rather than frequency invariance. Level fluctuations induced by spatial variance levels can lead to unreliable detection, as weaker signals might be missed or interpreted as noise, while stronger signals could be falsely identified as different sources. Optimizing for spatial invariance minimizes this risk since maintaining consistent levels helps in achieving a better SNR. A spatial consistent level, irregardless of direction, ensures that the signal stands out against the background noise, making it easier to identify and locate the target. By focusing on spatial invariance, the first deconvolution module 220 provides a stable and reliable foundation for subsequent signal analysis and interpretation.In a preferred embodiment, the first deconvolution module 220 addresses a first set of factors including at least one of: minimized spatial smearing due to beampatterns, increase in SNR and / or reduced frequency coloring. In another preferred embodiment, special emphasis is placed on reducing the spatial smearing of the conventional beamformer, meaning that in this preferred embodiment the first set of factors include at least Signal to Noise ratio. Optimizing the first deconvolution module 220 to address the spatial smearing effectively reduces the influence of ambient noise and increases the clarity and discernibility of the acoustic sources.In one alternative embodiment, the first deconvolution module 220 addresses a set of factors including at least White-Noise-Gain. The proposed first deconvolution module 220 may incorporates a neural network (NN) to significantly boost computational efficiency, especially in acoustic-based monitoring systems handling multiple objects. Traditional methods have limitations in dealing with non-linear distortions, varying environments, and simultaneous tracking of multiple objects. The proposed invention entailing; refining a global data set coarsely and subsequently selectively refining local parts of the acoustic data, achieves synergistic effects when combined with NNs for deconvolving acoustic data in large area monitoring.The identification module 300 is adapted to analyze the global acoustic image to identify the presence of potential objects of interest. The identification module initiates a location identification procedure if at least one object of interest is identified.The location identification procedure and the modules relevant for this are illustrated in figure 7, accompanying the relevant modules for the categorization procedure. The spatial orientation module 400 is adapted receive the global acoustic image and generate based thereon: a first spatial orientation data set and a second spatial orientation data; wherein said first spatial orientation data set comprises current elevation and azimuth angles for the at least one objects of interest; wherein said second spatial orientation data set comprises current apparent heading and angular speed for the at least one objects of interest, the second spatial orientation data set being based on a rate of change of elevation angle and / or azimuth angle for the at least one objects of interest. As can be seen in the figure, the spatial orientation module transmit the first and / or second spatial orientation data set to the local refinement module and may optionally also communicate this to the signature refinement module 700. Alternatively, the location module 600 may transmit the location data set to the signature refinement module 700 as to allow for a more precise local refinement performed by the signature refinement module 700.The local refinement module 500 is adapted to generate a locally refined acoustic image. The local refinement module 500 is adapted to refine the acoustic data or a refinement thereof only at sections which pertain to at least one object of interest. The sections which pertain to at least one object of interest being determined based on the first spatial orientation data set and / or the second spatial orientation data set;In one embodiment, the locally refined acoustic image generated by the local refinement module 600 is adapted to be used for localization of sound sources. In this embodiment, the local refinement module 600 preferably performs at least one of the following operations on the acoustic data only at sections which are estimated to pertain to the at least one object of interest:3. adjusting the phase and amplitude of the acoustic data or a refinement thereof using a first beamformer; and / or4. deconvolving the acoustic data or a refinement thereof using a second deconvolution module 520.The location module 600 may generate a categorization data set and / or a location data set based on the locally refined acoustic image. For use cases when both location and characteristics of the object are of importance, such as when monitoring for UAV detection or airplane collision prevention, the locally refined acoustic image generated by the local refinement module 500 may be tailored towards deriving location of sound sources, and a subsequent refinement may be made that is tailored towards categorization. However, even in such configurations the alternative to generate an initial categorization data set using the location module 600 could be advantageous. For example, if a categorization procedure is imitated subsequent to the location identification procedure, the initial categorization data set may comprise information advantageous to configure a second set of weighting factors and / or deconvolution in regards to a third set of factors used in a local refinement performed in the subsequent categorization procedure.In the illustrated embodiment, the location module 600 is adapted to generate a location data set based on the locally refined acoustic image wherein the location data set is indicating a location for the at least one objects of interest. As touched upon previously, the location module 600 may transmit the location data set to the signature refinement module 700.The illustrated embodiment shows a configuration where t the location module 600 is adapted for the specific purpose of localization and tracking of objects. In such an embodiment the second deconvolution module 520 is preferably optimized for spatial invariance rather than frequency invariance. Level fluctuations induced by spatial variance can lead to unreliable localization or tracking as weaker signals, or signals that change their relative distance or angle to the sensor might be missed or interpreted as noise, while stronger signals could be falsely identified as different sources. Optimizing for spatial invariance minimizes this risk since maintaining consistent levels helps in achieving a better SNR. A spatial consistent level, irregardless of direction, ensures that the signal stands out against the background noise, making it easier to identify and locate the target. By focusing on spatial invariance, the second deconvolution module 520 provides a stable and reliable foundation for subsequent signal analysis and interpretation.In one proposed embodiment, the second deconvolution module 520 addresses a second set of factors which relate to static source propagation effects and preferably comprises at least one of: Lioyd mirror effect, phantom sources and / or ground reflection.In the illustrated embodiment, a categorization procedure is initiated after at least one iteration of the location identification procedure is completed.If the categorization procedure is initiated, the categorization module 800 is adapted to generate a categorization data set based on the acoustic data 110 or a refinement thereof. The categorization data set comprises information relating to the characteristics of the at least one objects of interest.Preferably, the categorization module 800 communicates with or incorporates a signature refinement module 700 adapted to generate a signature refined acoustic image based on analyzing the acoustic data or a refinement thereof. The signature refinement module 700 is preferably refining only at sections which are estimated to pertain to the at least one object of interest. To determine which these sections are the signature refinement module 700 may use information received from the spatial orientation module 400 or the location module 600. The categorization data set could comprise any information relating to an object of interest which is relevant for the specific use case. For example, the information may include:12. an acoustic signature profile which can include parameters such as frequency spectrum characteristics, intensity patterns, temporal variations, and specific acoustic fingerprints that are unique to certain types of objects;13. spectral features such as Mel-frequency cepstral coefficients (MFCCs), spectral centroid, spectral bandwidth, spectral flatness, and / or spectral roll-off;14. temporal dynamics, meaning information indicating temporal evolution of the acoustic signal, which might include attack time, decay, sustain, release patterns, and temporal variations in intensity and frequency;15. Spatial characteristics representing the spatial distribution of the sound source, such as directionality, apparent size based on acoustic dispersion, and movement patterns based on changes in spatial attributes over time;16. Harmonic and percussive elements for identification of harmonic (tonal) and percussive (noisy / transient) components within the sound;17. Behavioral patterns, meaning information related to patterns of movement, speed, acceleration, and trajectory, which can be used to infer the nature of the object (e.g., vehicle, animal, human);18. Cross-referenced data, meaning information that correlates acoustic data with other types of sensor data, such as visual (from cameras), thermal, or radar, to enrich the categorization process.In the illustrated embodiment, the local refinement module 500 is adapted to adjust the acoustic data 110 or a refinement thereof using a first beamformer 510 and a second deconvolution module 520. Also, the signature refinement module 700 is adapted to adjust the acoustic data 110 or a refinement thereof using a second beamformer 710 and a second deconvolution module 720. In one preferred embodiment, at least one of: the first deconvolution module 220, the second deconvolution module 520, and / or the third deconvolution module 720 is using a neural network for deconvolving the acoustic data 110 or a refinement thereof with respect to at least one of the following artifacts: spatial smearing due to beampatterns, increase in SNR and / or reduced frequency coloring, atmospheric damping, ground reflection, doppler effect, Lloyd mirror effect and / or acoustic masking. Preferably the first deconvolving module 220 is using a recursive neural network. Also preferably, the first deconvolving module 220 is configured for deconvolving with respect to spatial smearing due to beampatterns, increase in SNR and / or reduced frequency coloring. While the second deconvolving module 520 preferably is configured for deconvolving with respect to Lioyd mirror effect, phantom sources and ground reflection. The third deconvolving module 720 is preferably configured for deconvolving with respect to doppler effect and atmospheric damping and frequency invariance.Each module within the illustrated system may be integrated with processing circuits, which may include microprocessors, digital signal processors (DSPs), and application-specific integrated circuits (ASICs). These circuits may be interfaced with computer-readable memories that store executable instructions for carrying out the module-specific functions.The processing circuits and memory units in each module are configured to receive, process, transmit, and store data in a coordinated manner, ensuring efficient operation of the overall monitoring system. The aforementioned modules perform operations which may involve advanced algorithms, including neural networks, embedded within the modules for processing and analyzing the acoustic data. As understood by the skilled person, the various modules mentioned in relation to the illustrated embodiment on figure 6 and 7 may be integrated or incorporated in a number of different configurations, meaning the function of one module, such as the global acoustic image module 200, may be performed within the same operation and using the same processing circuit which performs the function of another module, such as the identification module 300.In the context of the present application, acoustic data refers to the information captured from sound waves traveling through any medium, such as fluids (e.g. water), solids (e.g. earth), and gas (e.g. air). The sound waves, i.e., the oscillation in pressure, particle displacement, particle velocity and / or density, may sensed in a number of different ways depending on the given medium. What specific type of sensor is suitable depends largely on the given medium but could also be adapted for specific use cases. For example, when monitoring an area underwater a hydrophone array is suitable. If monitoring for objects moving in the air or on a surface on land, a microphone arrays is suitable. In avalanche or earthquake monitoring, i.e., when capturing sound waves moving through earth, a seismic sensor array implemented with geophone may be preferable. A seismic sensor might refer to any sensor capable of detecting movement within solids such as a load cell, laser doppler vibrometer and / or accelerometer array.In aspects, the proposed system according to aforementioned embodiment may be adapted specifically for aircraft surveillance. In this aspect the acoustic data is preferably acquired from at least one microphone array. In one embodiment of this aspect the system is specifically adapted for UAV surveillance, in other words drone detection. In other embodiments the system may survey for any aircraft which for the purpose of preventing collisions. In aspects, the proposed system according to aforementioned embodiment may be adapted specifically for geophysical mass flow monitoring. In this aspect the acoustic data is preferably acquired from at least one seismic sensor array. The geophysical mass flows monitored for may be objects such as avalanches, landslides or soil liquefaction. In aspects, the proposed system according to aforementioned embodiment may be adapted specifically for underwater monitoring. In this aspect the acoustic data is preferably acquired from at least one hydrophone array. In aspects, the proposed system according to aforementioned embodiment may be adapted specifically for land surface surveillance. In this aspect the acoustic data is preferably acquired from at least one microphone array. Land surface surveillance refers to monitoring for sound sources located on a surface of land, this may include application such as: studded tire monitoring, shooting / sniper detection, intrusion surveillance. It will be understood that the invention is not restricted to the aforedescribed and illustrated exemplifying embodiments thereof and that modifications can be made within the scope of the invention as defined by the accompanying claims.Enumerated itemized list of embodimentsItem 1. A method for monitoring an area using at least one acoustic sensor array, the method comprising the steps:A. acquiring acoustic data from the at least one acoustic sensor array; and acquiring environmental data comprising at least temperature in proximity to the at least one acoustic sensor array;B. processing the acoustic data and the environmental data, and acquiring a global acoustic image, wherein said processing comprises the operation:5. adjusting the phase and amplitude of the acoustic data using a delay and sum based beamformer;C. determining, based on at least the global acoustic image if there is at least one object of interest; and if at least one object of interest has been identified, initiating a location identification procedure; wherein the location identification procedure comprises at least the following steps:C,1. Estimating an initial location of the at least one object of interest based on the global acoustic data and either or both of:3. a set of elevation and azimuth angles for the at least one object of interest; and / or,4. a rate of change in elevation angle and azimuth angle for the at least one objects of interest; C,2. Processing the acoustic data and the environmental data or their respective refinements thereof, and acquiring a locally refined acoustic image which is refined only at sections which are estimated to pertain to the at least one object of interest;C,3. Processing at least the locally refined acoustic image, and acquiring a location data set and / or a categorization data set; wherein the location data set is indicating a location for the at least one objects of interest; and wherein the categorization data set comprises characteristics for the at least one object of interest. Item 2. The method according to item 1, wherein said processing in step B further comprises the operation:19. deconvolving the acoustic data with respect to a first set of factors influencing acoustic propagation characteristics using a neural network; preferably the first set of factors comprises at least one of: spatial smearing due to beampatterns, increase in SNR and / or reduced frequency coloring. Item 3. The method according to any previous item, wherein the processing in step C, 2. comprises performing at least one of the following operations on the acoustic data only at sections which are estimated to pertain to the at least one object of interest:5. adjusting the phase and amplitude of the acoustic data or a refinement thereof using a first set of weights and time delays; and / or6. deconvolving the acoustic data or a refinement thereof with respect to a second set of factors; wherein the first set of factors preferably relate to static source propagation effects and comprises at least one of: Lioyd mirror effect, phantom sources and / or ground reflection. Item 4. The method according to any previous item, wherein a categorization procedure is initiated after at least one iteration of the location identification procedure is completed, wherein the categorization procedure comprises the steps:D. processing the acoustic data or a refinement thereof at sections which are estimated to pertain to the at least one object of interest and acquiring a categorization data set indicates characteristics for the at least one object of interest; said processing comprising the steps:3. adjusting the phase and amplitude of the acoustic data or a refinement thereof using a second set of weights and time delays4. deconvolving the acoustic data or a refinement thereof with respect to a third set of factors; preferably the second set of artifacts relate to dynamic source propagation effects and / or comprises doppler effect and / or atmospheric damping. Item 5. The method according to any previous item, wherein at least one iteration of the categorization procedures is completed and wherein the categorization data set comprises an indication of a typical movement for the at least one of the object of interest; and wherein, in a next sequence of acoustic data, the estimation of the initial location of the at least one object of interest performed in step C, 2. is further based on the typical movement for the at least one of the object of interest. Item 6. A system for monitoring an area using at least one acoustic sensor array, wherein the system comprises:7. an information gathering module adapted to:5. receive acoustic data from the at least one acoustic sensor array; 6. receive environmental data from at least one ambient sensor and / or a secondary interface; wherein said environmental data comprises at least temperature in proximity to the at least one acoustic sensor array; 8. a global acoustic image module adapted to generate a global acoustic image based on the acoustic data and the environmental data, wherein the global acoustic image is generated by:7. acquiring a phased array output by using a delay and sum based beamformer module; wherein the delay and sum based beamformer module is adapted to adjust the phase and amplitude of the acoustic data;8. acquiring the global acoustic image by deconvolving the phased array output using a first deconvolution module;9. a identification module adapted to analyze the global acoustic image to identify the presence of potential objects of interest, wherein the identification module initiates a location identification procedure if at least one object of interest is identified;10. a spatial orientation module adapted receive the global acoustic image and generate based thereon: a first spatial orientation data set and a second spatial orientation data; wherein said first spatial orientation data set comprises current elevation and azimuth angles for the at least one objects of interest; wherein said second spatial orientation data set comprises current apparent heading and angular speed for the at least one objects of interest, the second spatial orientation data set being based on a rate of change of elevation angle and / or azimuth angle for the at least one objects of interest;11. a local refinement module adapted to generate a locally refined acoustic image; wherein the local refinement module is adapted to refine the acoustic data or a refinement thereof only at sections which pertain to at least one object of interest; said sections being determined based on the first spatial orientation data set and / or the second spatial orientation data set;12. a location module set adapted to generate a location data set and / or a categorization data set based on the locally refined acoustic image wherein the second location data set is indicating a location for the at least one objects of interest, and wherein the categorization data set comprises characteristics for the at least one object of interest. Item 7. The system according to item 6, wherein a categorization procedure is initiated after at least one iteration of the location identification procedure is completed, wherein the categorization procedure comprises a categorization module adapted to generate a categorization data set based on the acoustic data or a refinement thereof; wherein the categorization data comprises information relating to the characteristics of the at least one objects of interest. Item 8. The system according to item 7, wherein the system comprises a signature refinement module adapted to generate a signature refined acoustic image based on analyzing the acoustic data or a refinement thereof; wherein the signature refinement module is refining only a part of the acoustic data or a refinement thereof based at least on the first spatial orientation data set and / or the second spatial orientation data set. Item 9. According to any of item 6 to 8, wherein the local refinement module is adapted to adjust the acoustic data or a refinement thereof using a first beamformer and a second deconvolution module; wherein the signature refinement module is adapted to adjust the acoustic data or a refinement thereof using a second beamformer and a third deconvolution module; wherein at least one of the following deconvolving modules: the first deconvolving module; the second deconvolving module; and / or the third deconvolving module, is using at least one neural network for deconvolving the acoustic data or a refinement thereof with respect to at least one of the following artifacts: atmospheric damping, ground reflection, doppler effect, Lloyd mirror effect and / or acoustic masking; preferably the first deconvolving module is configured for deconvolving with respect to White-Noice-Gain (WNG), frequency coloring, as well as spatial and time resolution; preferably the second deconvolving module is configured for deconvolving with respect to Lioyd mirror effect, phantom sources and ground reflection; preferably the third deconvolving module is configured for deconvolving with respect to doppler effect and atmospheric damping.

Claims

1. A method performed by at least one processing circuit for monitoring an area using at least one acoustic sensor array, the method comprising the steps:A. acquiring acoustic data from the at least one acoustic sensor array; and acquiring environmental data comprising at least temperature in proximity to the at least one acoustic sensor array, the environmental data being indicative of how the environment in the monitored area influence acoustic propagation characteristics;B. processing the acoustic data and the environmental data to acquire a global acoustic image, wherein said processing comprises the operation:- adjusting the phase and amplitude of the acoustic data using a delay and sum based beamformer;C. determining, based on at least the global acoustic image if there is at least one object of interest; and if at least one object of interest has been identified, initiating a location identification procedure; wherein the location identification procedure comprises at least the following steps:C,1. Estimating an initial location of the at least one object of interest based on the global acoustic image and either or both of:- a set of elevation and azimuth angles for the at least one object of interest derived from the global acoustic image; and,- a rate of change in elevation angle and azimuth angle for the at least one objects of interest derived from the global acoustic image; C, 2. Processing the acoustic data and the environmental data or processing the global acoustic image using a neural network to acquire a locally refined acoustic image which is refined only at sections which are estimated to pertain to the at least one object of interest;C,3. Processing at least the locally refined acoustic image, and acquiring a location data set and / or a categorization data set; wherein the location data set is indicating a location for the at least one objects of interest; and wherein the categorization data set comprises characteristics for the at least one object of interest. 2. The method according to claim 1, wherein said processing in step B further comprises the operation:deconvolving the acoustic data with respect to a first set of factors influencing acoustic propagation characteristics using a neural network; preferably the first set of factors comprises at least one of: spatial smearing due to beampatterns, increase in SNR and / or reduced frequency coloring. 3. The method according to any previous claim, further comprising performing at least one of the following operations on the acoustic data only at sections which are estimated to pertain to the at least one object of interest, such that the locally refined acoustic image has a higher resolution than the global acoustic image:- adjusting the phase and amplitude of the acoustic data using a first set of weights and time delays; and / or- deconvolving the acoustic data with respect to a second set of factors; wherein the second set of factors preferably relate to static source propagation effects and comprises at least one of: Lloyd mirror effect, phantom sources and / or ground reflection. 4. The method according to any previous claim, wherein a categorization procedure is initiated after at least one iteration of the location identification procedure is completed, wherein the categorization procedure comprises the steps:D. processing the acoustic data thereof at sections which are estimated to pertain to the at least one object of interest and acquiring a categorization data set indicates characteristics for the at least one object of interest; said processing comprising the steps:- adjusting the phase and amplitude of the acoustic data using a second set of weights and time delays- deconvolving the acoustic data with respect to a third set of factors; preferably the second set of artifacts relate to dynamic source propagation effects and / or comprises doppler effect and / or atmospheric damping. 5. The method according to any previous claim, wherein at least one iteration of the categorization procedures is completed and wherein the categorization data set comprises an indication of a typical movement for the at least one of the object of interest; and wherein, in a next sequence of acoustic data, the estimation of the initial location of the at least one object of interest performed in step C, 1. is further based on the typical movement for the at least one of the object of interest. 6. A system for monitoring an area using at least one acoustic sensor array, wherein the system comprises:- an information gathering module adapted to:- receive acoustic data from the at least one acoustic sensor array; - receive environmental data from at least one ambient sensor and / or a secondary interface; wherein said environmental data comprises at least temperature in proximity to the at least one acoustic sensor array, said environmental data being indicative of how the environment in the monitored area influence acoustic propagation characteristics; - a global acoustic image module adapted to generate a global acoustic image based on processing the acoustic data and the environmental data, wherein the said processing to generate the global acoustic image comprises:- acquiring a phased array output by using a delay and sum based beamformer module; wherein the delay and sum based beamformer module is adapted to adjust the phase and amplitude of the acoustic data;- acquiring the global acoustic image by deconvolving the phased array output using a first deconvolution module;- an identification module adapted to analyze the global acoustic image to identify the presence of potential objects of interest, wherein the identification module initiates a location identification procedure if at least one object of interest is identified;- a spatial orientation module adapted receive the global acoustic image and generate based thereon: a first spatial orientation data set and a second spatial orientation data; wherein said first spatial orientation data set comprises current elevation and azimuth angles for the at least one objects of interest; wherein said second spatial orientation data set comprises current apparent heading and angular speed for the at least one objects of interest, the second spatial orientation data set being based on a rate of change of elevation angle and / or azimuth angle for the at least one objects of interest;- a local refinement module adapted to process the acoustic data and the environmental data or process the global acoustic image using a neural network to generate a locally refined acoustic image which is refined only at sections which pertain to at least one object of interest; said sections being determined based on the first spatial orientation data set and / or the second spatial orientation data set;- a location module set adapted to generate a location data set and / or a categorization data set based on the locally refined acoustic image wherein the location data set is indicating a location for the at least one objects of interest, and wherein the categorization data set comprises characteristics for the at least one object of interest. 7. The system according to claim 6, wherein a categorization procedure is initiated after at least one iteration of the location identification procedure is completed, wherein the categorization procedure comprises a categorization module adapted to generate a categorization data set based on the acoustic data; wherein the categorization data comprises information relating to the characteristics of the at least one objects of interest.  8. The system according to claim 7, wherein the system comprises a signature refinement module adapted to generate a signature refined acoustic image based on analyzing the acoustic data; wherein the signature refinement module is refining only a part of the acoustic data based at least on the first spatial orientation data set and / or the second spatial orientation data set. 9. The system according to claim 8, wherein the local refinement module is adapted to adjust the acoustic data using a first beamformer and a second deconvolution module such that the locally refined acoustic image has a higher resolution than the global acoustic image; wherein the signature refinement module is adapted to adjust the acoustic data using a second beamformer and a third deconvolution module; wherein at least one of the following deconvolving modules: the first deconvolving module; the second deconvolving module; and / or the third deconvolving module, is using at least one neural network for deconvolving the acoustic data with respect to at least one of the following artifacts: atmospheric damping, ground reflection, doppler effect, Lloyd mirror effect and / or acoustic masking; preferably the first deconvolving module is configured for deconvolving with respect to White-Noise-Gain (WNG), frequency coloring, as well as spatial and time resolution; preferably the second deconvolving module is configured for deconvolving with respect to Lloyd mirror effect, phantom sources and ground reflection; preferably the third deconvolving module is configured for deconvolving with respect to doppler effect and atmospheric damping.  

Images