A PROCESS FOR INTELLIGENT SOUND CLASSIFICATION USING DEVICES AND ALGORITHMS FOR ECHOACOUSTIC INTERVENTION BASED ON REAL-TIME HUMAN PERCEPTION

Description

/ 12 PROCESS FOR INTELLIGENT SOUND CLASSIFICATION USING DEVICES AND ALGORITHMS FOR ECHOACOUSTIC INTERVENTION Based on real-time human perception. Field of Invention

[001] G10K 11 / 16 Methods or devices for protecting against, or for damping, noise or other acoustic waves in general. Fundamentals of the invention and state of the art

[002] The process for intelligent sound classification using devices and algorithms for echoacoustic intervention based on real-time human perception describes a process for automating the analysis, classification, and intervention in the soundscape, based on user perception.

[003] The process allows the entire process of capturing, processing, analyzing, and ordering the intervention in the soundscape to be carried out in a single device, based on a subjective classification of sensations or feelings.

[004] Accurate characterization of sound environments is fundamental for the effective implementation of noise control measures. Acoustic analysis, using specialized instrumentation and signal processing software, allows for the quantification and qualification of sound pressure levels, frequency spectrum, reverberation time, and other relevant parameters. Techniques such as individual dosimetry, spectral analysis, and acoustic mapping are widely used to assess human exposure to noise and identify the main sound sources.

[005] Environmental noise analysis, while crucial for acoustic quality management, faces several challenges when performed exclusively by people. Human perception of sound is subjective and influenced by factors Petition 870250074313, dated 08 / 22 / 2025, page 3 / 22 / 12, such as individual hearing threshold, auditory adaptation, masking effect, and psychological factors. Furthermore, the temporal variability of noise levels makes it difficult to obtain precise and representative measurements.

[006] To overcome these limitations, acoustic instrumentation, such as sound level meters and spectrum analyzers, offers greater objectivity, repeatability, and the ability to perform detailed analysis.

[007] However, combining subjective and objective methods, through questionnaires and interviews, along with instrumental analysis, becomes a complex and slow process, as it requires different stages and instruments. Furthermore, current methods do not allow for real-time intervention in the soundscape, based on the user's feelings and perceptions.

[008] The process for intelligent sound classification using devices and algorithms for ecoacoustic intervention based on real-time human perception combines aspects of user subjectivity in the classification of the acoustic landscape, being able to perform analysis, classification and decision support in a single device.

[009] Currently, it is not possible to classify the soundscape automatically in real time based on feelings. For example, in the methodology described in the ISO / TS 12913-3:2019 standard, an environment can be classified according to the user's feelings and perceptions, such as calm, boring, stressful or vibrant, but this is done based on interviews followed by data analysis, using statistical and mathematical methods, which does not allow for real-time classification of the acoustic landscape.

[010] Current methods for classifying the soundscape for animals are subjective and based on the perception of the owner or breeder, not allowing for standardized monitoring of echoacoustic quality beyond sound pressure levels or noise. Petition 870250074313, dated 08 / 22 / 2025, page 4 / 22 / 12

[011] Current processes involve the execution of multiple steps that include data collection, decoding, and sending for processing in software for subsequent decision-making. Current processes require a combination of various equipment, for example, sound level meters, audio recorders, storage media, audio processing software, software for spectrogram analysis, and human interpretation.

[012] To classify the soundscape, based on artificial intelligence methods, it is necessary to include machine learning steps after processing the audio data and acoustic spectra. This procedure is not performed on the same sound capture equipment or even on a portable device that allows the user to calibrate the equipment according to their emotions.

[013] In the current state of the art, it is not possible to perform all these operations involving the classification of the acoustic landscape, based on human perceptions, to generate data that enable decision-making.

[014] The lack of real-time analysis makes it difficult to make decisions regarding soundscape intervention. Existing soundscape intervention processes are basically based on controlling the acoustic pressure of a given piece of equipment, when it has volume control, for example. If it is not possible to control the volume of a given piece of equipment, it is not possible for the user to assess the quality of the soundscape based on their emotional preferences.

[015] Another problem related to soundscape intervention lies in the decision-making process regarding whether the user should leave or whether a third person or other living being affected by the soundscape should be removed. For example, in cases of inadequate soundscapes or those that may cause discomfort to children, the lack of a process that allows the responsible party to make a decision about whether or not to remain in the environment can jeopardize the health of children. Petition 870250074313, dated 08 / 22 / 2025, page 5 / 22 / 12

[016] In breeding, care, or other circumstances that demand high-quality echoacoustic conditions for animals, there is no intelligent process capable of assisting in decision-making about the sound environment, based on subtle behavioral reactions of the specimens. Current systems are based primarily on noise measurements or sound spectrograms that do not allow for intelligent calibration of the relationship between behavior and soundscape.

[017] Accurate characterization of sound environments is essential for implementing effective noise control measures, but this task is far from simple. It involves several technical challenges that require both specialized knowledge and sensitivity to human perceptions.

[018] Among the technical problems are the need to use sophisticated equipment, such as sound level meters and spectrum analyzers, which are expensive and complex to operate for users without specialized training. These devices are capable of accurately measuring parameters such as sound pressure levels and frequency spectrum. However, in complex environments, such as urban areas full of traffic, isolating different noise sources to obtain clear measurements can be very difficult. Operating this equipment and interpreting the data correctly requires experience and specialization.

[019] Another problem lies in the subjectivity of human sound perception. Each person perceives sound differently, influenced by factors such as auditory sensitivity and emotional state. For example, in a park, birdsong may be relaxing for some and annoying for others. This diversity of perceptions makes it difficult to create solutions that satisfactorily meet everyone's needs.

[020] The variability of noise over time and in different locations is another difficult aspect to deal with. In a factory, for example, noise levels can Petition 870250074313, dated 08 / 22 / 2025, page 6 / 22 / 12 vary greatly depending on the time of day or the specific area. Capturing this variation requires continuous measurements at multiple points, which makes the analysis more complex and laborious.

[021] The combination of objective and subjective analysis methods also constitutes a technical problem, as it requires an interface that interacts with the user by matching echoacoustic data with their perceptions simultaneously. With current methods, perception is obtained through questionnaires and interviews. Finding a balance between these two aspects is crucial to creating sound environments that are both technically effective and satisfying for people.

[022] The lack of uniformity in the application of these methodologies is another relevant technical problem. Different projects may adopt varied approaches, which makes it difficult to compare results. Without clear standardization, it becomes difficult to ensure that noise control measures are consistent and effective in different contexts.

[023] Another challenge is communicating the results. Acoustic measurements and analyses produce complex data that are often difficult to interpret for people who are not trained in the field. Summary of the invention

[024] The process for intelligent sound classification using devices and algorithms for echoacoustic intervention based on real-time human perception simplifies the analysis and presentation of perceptual classification results to the user. In addition, the process allows information to be sent to external devices in order to adapt the environment to a desired condition in an automated way.

[025] There are basically processes and products available on the market that allow the user to adjust the sound pressure (volume) or customize it in a way Petition 870250074313, dated 08 / 22 / 2025, page 7 / 22 / 12 manual the bass and treble settings originating from a specific sound source, generally restricted to music. Furthermore, perception, which is a subjective scale based on human feelings, is not considered.

[026] The process for unsupervised sound classification for echoacoustic intervention based on human perception algorithms allows a user or group of users to record the classification of the acoustic quality of the environment, generating a classification algorithm that subsequently classifies the acoustic environment in an unsupervised manner. The classification data can be sent to the user to, for example, issue an alert or activate some acoustic control device aimed at intervention in the environment.

[027] Among the possible, but not limited to, applications are the monitoring and control of acoustic and echoacoustic quality in environments for people with autism spectrum disorder (ASD), attention deficit disorder (ADD), attention deficit hyperactivity disorder (ADHD), hearing impairment, visual impairment, hypersensitivity to sounds, among other situations where environmental acoustic quality is desired.

[028] It also allows for the creation of a quality acoustic environment aimed at environmental comfort, promoting well-being, preventing health problems and other aspects that the user deems beneficial.

[029] It can also be applied to animal welfare based on the classification of behavioral attributes, both for those living in natural and artificial environments.

[030] The process allows the integration of acoustic environment information with other environmental variables, such as air temperature, wind speed, relative humidity, illuminance, among others. Petition 870250074313, dated 08 / 22 / 2025, page 8 / 22 / 12

[031] The process can be performed by combining different electronic devices, with dedicated or non-dedicated hardware and software, mobile or fixed, embedded and non-embedded, installed in different structures with diverse dimensions, such as masonry, vehicles, aircraft, auditoriums, gymnasiums, pastures, crops and natural areas.

[032] The data generated in the process are georeferenced and incorporated into a geographic information system (GIS). Description of the figures

[033] The present inventive activity will be described in greater detail below, by way of non-limiting example, with reference to its preferred embodiment illustrated in the drawings below, in which: Figure 1 shows the process for intelligent sound classification using devices and algorithms for echoacoustic intervention based on real-time human perception. Figure 2 shows Step (11) of the machine learning process for intelligent sound classification using devices and algorithms for echoacoustic intervention based on real-time human perception. Figure 3 presents the interface options for steps (10), (20) and (24) for the intelligent sound classification process using devices and algorithms for echoacoustic intervention based on real-time human perception. The interfaces are customizable for each desired application. Figure 4 presents a possible example of a decision tree in a residence for the intelligent sound classification process using devices and algorithms for echoacoustic intervention based on real-time human perception. Petition 870250074313, dated 08 / 22 / 2025, page 9 / 22 / 12 Figure 5 presents possible examples of decision trees in animal breeding for the intelligent sound classification process using devices and algorithms for echoacoustic intervention based on real-time human perception. Figure 6 presents possible examples of decision trees in open and residential areas for the intelligent sound classification process using devices and algorithms for echoacoustic intervention based on real-time human perception. Figure 7 illustrates an example of the external installation options for the unsupervised sound classification process for echoacoustic intervention based on human perception algorithms. Various types of acoustic sensors can be adopted, installed according to user requirements. Detailed description of the invention

[034] The process for intelligent sound classification using devices and algorithms for ecoacoustic intervention based on real-time human perception consists of classifying the acoustic environment according to the user(s) perception and generating an automated classification of acoustic quality based on their predefined preferences.

[035] Input material consists of sounds or vibrations from any source, a user's subjective perception, and environmental data. Output consists of classification data that can be displayed to the user for decision-making or used for automation and control of acoustic manipulation systems.

[036] The entire process is developed using a device or set of devices, controlled by interfaces. The devices and interfaces together control the tasks of capturing acoustic signals, environmental data, Petition 870250074313, dated 08 / 22 / 2025, page 10 / 22 / 12 user perception, data analysis, display of results and data output for automation.

[037] The process is performed in two stages. Phase 1 consists of calibrating the model based on the user's perception of the quality of the acoustic environment. Phase 2 involves analyzing the data in real time and expressing the results. The general structure of the process is described in Figure 1.

[38] For calibration (Phase 1) The process begins with the initial capture of audio signals and environmental data through specialized sensors (1). Next, the captured audio signals and environmental data are transmitted to subsequent processing modules (2), followed by continued transmission and preparation of signals and data for processing (3). The captured audio signals are then sent to the processing modules (4), and subsequently stored in digital format for later analysis (5). The initial analysis of the audio signals identifies characteristics of the soundscape (6), followed by processing of the audio signals and environmental data to extract relevant metrics (7). Machine learning algorithms are applied for classification and interpretation of the extracted data (8). The user's perception of the sound environment is collected and processed by a specific interface (9).The system has an interface module that allows interaction with the user and visualization of the processed data (10), while the classified data is stored after classification performed by machine learning algorithms (11).

[039] For the real-time analysis stage (Phase 2), dedicated sensors continuously capture environmental data (12) together with soundscape data (14), which are processed in the same way as in Phase 1. The processed environmental and audio data are compared with reference algorithms for evaluation (13) in a real-time comparison process (15). An interface displays the processed results to the user, including metrics and classifications (16), and a module calculates and displays the accuracy level of the Petition 870250074313, dated 08 / 22 / 2025, page 11 / 22 / 12 classifications performed by the system (17). A decision module determines subsequent actions based on the processed data and comparative results (18). The process also integrates external equipment for additional data collection or processing (19). The final interface consolidates all interactions and presents the final result to the user (20). The accuracy of the classification is evaluated by the user (21) and sent to adjust the calibration of the models (Phase 1) by machine learning (9, 10 and 11). The final decision based on the processed data is presented to the user for action or information (22). Data provided by external equipment is stored and used as part of the decision process (23). An interface manages the integration between the main system and external equipment or sensors (24).Finally, the process is completed with the display of all results and metrics to the user, ready for reports or subsequent actions in real time (25).

[040] For the metrics extraction step (8), soundscape index calculations are performed, based on Table 1. This step is performed with the support of software installed on the device, the metrics results are combined with user perception (9) using the human interface (10). The metrics can also be combined with each other to generate or create new metrics, based on spectral analysis procedures. Table 1: Indices calculated in the extraction phase of the metrics (8) of the audio signals of the process for intelligent sound classification using devices and algorithms for echoacoustic intervention based on real-time human perception. Index Principle Mathematical Formula Reference Temporal Entropy Index Complexity of the temporal envelope Ht = -Z(p(t) * log(p(t))) [1] Spectral Entropy Index Complexity of the frequency spectrum Hf = -Z(p(f) * log(p(f))) [1] Petition 870250074313, dated 08 / 22 / 2025, page 12 / 22 / 12 Acoustic Entropy Index Complexity of the temporal envelope and frequency spectrum H = Ht * Hf [1] Acoustic Complexity Index Complexity of the spectrogram ACI = Z|X(t) - X(t+1)| / £X(t) [2] Acoustic Richness Index Complexity of the temporal envelope and intensity AR = (Ht * I) / Hf [3] Acoustic Diversity Index Complexity of the frequency spectrum ADI = Hf / Σ(ρ(^Λ2) [4] Normalized Soundscape Difference Index Ratio between anthropony and biophony NDSI = (BA) / (B+A) [5]

[041] In the machine learning stage (11), several models are adjusted in order to obtain the best possible classification of user perceptions (9) with acoustic landscape metrics (8), as described in Figure 2.

[042] The steps that make up the use of human interface (10), (20) and (24) can be performed using applications on mobile devices. In this scenario, all steps of the process can be performed on a single mobile device.

[043] Interfaces are essential for calibration procedures, display of results and validation through user evaluation of the result.

[044] The interfaces allow geographic information systems (GIS) to be incorporated into soundscape data, enabling, for example, decisions about staying in a particular location or searching for locations with better soundscape conditions. The interfaces also allow for various customizations regarding applications and analysis scenarios.

[045] For step (20) several decision trees can be programmed according to the applications of the process described. The programming of one or Petition 870250074313, dated 08 / 22 / 2025, page 13 / 22 / 12 more decision trees depend on the applications and expectations of use of the results by the user or set of users. Figures 4, 5 and 6 present possible examples of decision trees in different scenarios. Petition 870250074313, dated 08 / 22 / 2025, page 14 / 22

Claims

1 / 3 CLAIMS [1] “PROCESS FOR INTELLIGENT SOUND CLASSIFICATION USING DEVICES AND ALGORITHMS FOR ECOACOUSTIC INTERVENTION BASED ON REAL-TIME HUMAN PERCEPTION”, characterized by enabling a single electronic device to translate human feelings and perceptions induced by the acoustic environment, environmental data into algorithms for generating information for the user and automating audio devices; [2] “PROCESS FOR INTELLIGENT SOUND CLASSIFICATION USING DEVICES AND ALGORITHMS FOR ECOACOUSTIC INTERVENTION BASED ON REAL-TIME HUMAN PERCEPTION”, according to claim 1, is characterized by allowing a user or set of users to record the classification of the acoustic quality of the environment, generating a classification algorithm that subsequently classifies the acoustic environment in an unsupervised manner; [3] “PROCESS FOR INTELLIGENT SOUND CLASSIFICATION USING DEVICES AND ALGORITHMS FOR ECOACOUSTIC INTERVENTION BASED ON REAL-TIME HUMAN PERCEPTION”, according to any of the preceding claims, is characterized by allowing the integration of acoustic environment information with other environmental variables, such as air temperature, wind speed, relative air humidity, illuminance, among others, aiming at a broader characterization of the soundscape; [4] “PROCESS FOR INTELLIGENT SOUND CLASSIFICATION USING DEVICES AND ALGORITHMS FOR ECOACOUSTIC INTERVENTION BASED ON REAL-TIME HUMAN PERCEPTION”, according to any of the preceding claims, is characterized by allowing classification data to be sent to the user to issue an alert or trigger some acoustic control device, aiming at an action or intervention in the environment; Petition 870250014595, dated 23 / 02 / 2025, p. 18 / 28 2 / 3 [5] “PROCESS FOR INTELLIGENT SOUND CLASSIFICATION USING DEVICES AND ALGORITHMS FOR ECOACOUSTIC INTERVENTION BASED ON REAL-TIME HUMAN PERCEPTION”, according to any of the preceding claims, is characterized by enabling the improvement of environmental conditions for people, with possible, but not limited to, applications such as monitoring and controlling acoustic and echoacoustic quality in environments for people with autism spectrum disorder (ASD), attention deficit disorder (ADD), attention deficit hyperactivity disorder (ADHD), hearing impairment, visual impairment, hypersensitivity to sounds, among other situations where environmental acoustic quality is desired; [6] “PROCESS FOR INTELLIGENT SOUND CLASSIFICATION USING DEVICES AND ALGORITHMS FOR ECOACOUSTIC INTERVENTION BASED ON REAL-TIME HUMAN PERCEPTION”, according to any of the previous claims, is characterized by enabling the creation of a quality acoustic environment aimed at environmental comfort, promoting well-being, preventing health problems and other aspects that the user deems beneficial; [7] “PROCESS FOR INTELLIGENT SOUND CLASSIFICATION USING DEVICES AND ALGORITHMS FOR ECOACOUSTIC INTERVENTION BASED ON REAL-TIME HUMAN PERCEPTION”, according to any of the preceding claims, is characterized by being applicable to animal welfare based on the classification of behavioral attributes, both for those living in natural and artificial environments; [8] “PROCESS FOR INTELLIGENT SOUND CLASSIFICATION USING DEVICES AND ALGORITHMS FOR ECOACOUSTIC INTERVENTION BASED ON REAL-TIME HUMAN PERCEPTION”, according to any of the preceding claims, is characterized by being able to be performed by combining different electronic devices, with dedicated or non-dedicated hardware and software, mobile or fixed, embedded and not embedded, installed in different structures with different dimensions, such as masonry, vehicles, aircraft, auditoriums, gymnasiums, pastures, crops and natural areas; [9] “PROCESS FOR INTELLIGENT SOUND CLASSIFICATION USING DEVICES AND ALGORITHMS FOR ECOACOUSTIC INTERVENTION BASED ON REAL-TIME HUMAN PERCEPTION”, according to any of the previous claims, is characterized by being scalable, and can be adopted in closed or open environments of different dimensions or shapes. Petition 870250014595, dated 23 / 02 / 2025, p. 20 / 28

Images