Sound-signal analysis system and method for automatic estimation of uroflowmetry parameters with real-time frequency volume chart integration and automated voiding pattern detection including nocturia parameter analysis

Abstract

The present invention relates to a system for analyzing an audio signal to evaluate urinary flow rate, comprising an audio signal receiving unit configured to receive and record urination sound from a recording device; a signal processing and intelligent model unit configured to preprocess the audio signal, extract audio features, and input feature values to a machine-learning model to estimate urinary flow parameters; a display and reporting unit configured to present urinary flow parameters versus time; and a database and connectivity module configured to transmit and store results in a central database or cloud system and to connect with an external device or service. The system further derives a nocturia parameter by identifying nocturnal voiding events and aggregating corresponding flow parameters. The invention also discloses a method for analyzing an audio signal using the system.

Description

[0001] SOUND-SIGNAL ANALYSIS SYSTEM AND METHOD FOR AUTOMATIC ESTIMATION OF UROFLOWMETRY PARAMETERS WITH REAL-TIME FREQUENCY VOLUME CHART INTEGRATION AND AUTOMATED VOIDING PATTERN DETECTION INCLUDING NOCTURIA PARAMETER ANALYSIS

[0002] Technical Field

[0003] The present invention relates to the field of engineering and digital health technology, and more particularly to a system and a method for analyzing an audio signal to evaluate a urinary flow rate.

[0004] Background Art

[0005] From medical information, it is found that Lower Urinary Tract Symptoms (LUTS) are commonly found in the adult population and tend to increase with age (Irwin et al., 2006; Liu et al., 2019). In general, evaluation of such symptoms uses uroflowmetry, which is a noninvasive method for analyzing a patient's voiding function.

[0006] The clinical practice recommendation of the European Association of Urology (EAU) indicates that uroflowmetry is used as part of the initial evaluation for both male and female patients with LUTS (Cornu et al., 2024). If the maximum flow rate (Qmax) is lower than 10 milliliters per second, it often indicates urinary tract obstruction or reduced flow, which is at risk of acute urinary retention, particularly in male patients with Benign Prostatic Hyperplasia (BPH) (Cornu et al., 2024; Summers et al., 2021). However, performing uroflowmetry in a healthcare facility may have limitations in terms of time, insufficient voided volume, and patient anxiety and stress during the test, causing the test results to be inaccurate (Almas et al., 2024). Therefore, new technological innovations have been developed to replace or enhance the efficiency of traditional testing.

[0007] One studied method is sound-based uroflowmetry, in which urination sound signals are processed into parameter values comparable to standard uroflowmetry (Alvarez et al., 2024). There are studies showing consistency between measured values and the traditional system, such as maximum flow rate, average flow rate (Qave), and voided volume, and the method can also classify abnormal flow patterns (Lee et al., 2023).

[0008] In developing such technology, various research groups have applied deep learning and audio signal analysis, such as Mel-Frequency Cepstral Coefficients (MFCC) and classification using Support Vector Machines (SVM) (Bahatti, 2016), to improve the accuracy androbustness of predicting urinary flow parameters from sound. Although some studies indicate clinical feasibility, further development is still required in accuracy, personalization, and practicality of deployment.

[0009] Currently, uroflowmetry is an important method for evaluating abnormalities of the lower urinary tract or various voiding problems. However, most conventional testing must be performed in a healthcare facility, which may cause inconvenience, such as waiting for an appointment, and at the time of testing the voided volume may be insufficient or there may be privacy concerns, resulting in inaccurate results or reduced willingness to attend frequently. In crowded hospital situations, these processes become slower and patients may cancel the test.

[0010] Patent Literature

[0011] PTL1: EP4212096A1;

[0012] PTL2: US11604177B1;

[0013] PTL3: US20220378348A1;

[0014] PTL4: US20210121112A1.

[0015] A search of international patent databases indicates inventions related to urinary flow measurement systems in various forms.

[0016] For example, European patent application publication EP4212096A1 discloses a double-container structure (inner and outer container) with capacitive electrodes to measure liquid level together with motion and temperature sensors, and may optionally include an acoustic sensor or a spectrophotometric device for specialized measurements.

[0017] U.S. granted patent US11604177B1 presents a smart toilet having multiple modules such as a urinalysis module, a stool analysis module, an in-bowl camera, and may include a uroflowmetry module and a user identification module, designed for integrated home health monitoring.

[0018] U.S. patent application publication US20220378348A1 describes a portable flow-rate measurement device that can store data and link to an automatic voiding diary system.

[0019] U.S. patent application publication US20210121112A1 presents a wearable flow-rate measurement form factor with sensors installed within a urine funnel, considering convenience and accuracy in measuring various parameters. Although the above innovations show progress in measuring urinary flow rate in various contexts — portable, wearable, or integrated with a smart toilet — most still require specialized hardware or operation in a healthcare facility, or have complexity that limits accessibility due to cost, equipment, or healthcare infrastructure.Upon reviewing Thailand's patent databases, no invention is found that corresponds or is similar to the use of Artificial Intelligence (Al) to predict a urinary flow rate from an audio signal in a manner designed for convenient use with automated processing, including linking Internet of Things (loT) and Internet of Medical Things (loMT) together with Big Data to form an intelligent health system for comprehensive diagnosis and follow-up of urinary tract conditions. No invention in Thailand is found to have a comparable structure or key elements sufficiently identical or similar to be considered the same work or capable of imitation. Therefore, there arises a concept of developing an invention for measuring urinary flow rate through urination sound signals that can be installed or used in daily life by general users, particularly using Al to analyze urination sound signals to compute parameters comparable to standard uroflowmetry and to securely and rapidly link the analysis results into a central database. This leads to improved preventive healthcare services. Users can check their voiding condition anytime and anywhere, reducing waiting time, congestion, and limitations in access to testing in hospitals, which is consistent with current trends in digital health and telemedicine that emphasize convenient access to health services through technology.

[0020] Summary of Invention

[0021] In summary, the present invention not only aims to solve the limitations of conventional in-hospital uroflowmetry systems, but also extends to a platform in which patients or users can use Al to analyze urination sounds by themselves and link to Big Data so that medical personnel or researchers can monitor results promptly, supporting diagnosis and treatment effectively, and promoting self-care without the need to travel to a hospital each time uroflowmetry is required.

[0022] Technical Problem

[0023] In general clinical evaluation of voiding function, it is necessary to use a uroflowmeter installed in a healthcare facility and operated by medical personnel, resulting in inconvenience for patients and making it difficult to monitor symptoms in a real-life living environment.

[0024] Solution to Problem

[0025] To address the foregoing, a system and a method for analyzing an audio signal to evaluate a urinary flow rate (FLOWMIND-RA) are developed such that a user can use a smartphone or a general audio recording device to record urination sound. The system then performs automated and wireless processing to calculate urinary flow parameters, such as maximum flow rate (Qmax), average flow rate (Qave), and voided volume (Voided Volume),in a timely manner, thereby supporting screening and follow-up of urinary tract abnormalities anytime and anywhere, reducing travel burden and increasing follow-up frequency.

[0026] In the invention, the user records sound via a smartphone or a general recording device, and an audio input module forwards the data to a signal processing and intelligent model unit (Signal Processing & Al Engine) to perform preprocessing and feature extraction, for example Mel- Frequency Cepstral Coefficients (MFCC) and Zero Crossing Rate (ZCR), prior to estimating uroflowmetry parameters including maximum flow rate (Qmax), average flow rate (Qave), and voided volume (Voided Volume). The results undergo post-processing and are presented in a user interface and reporting module (User Interface & Reporting) (13), and are connected to and stored via a database / IoT / IoMT integration module (Database / IoT / IoMT Integration) (12). The system architecture is divided into a frontend layer, a backend and business logic layer (Backend & Business Logic Layer) (31), a machine learning pipeline, and a database. The operational steps correspond to flowcharts (20)-(27). The invention aims to enable convenient, reliable, and continuous self-screening and monitoring of urinary tract conditions outside a healthcare facility.

[0027] The processing workflow comprises:

[0028] (i) an element and key-feature analysis step including signal acquisition and preparation for feature extraction, wherein the audio input module receives / records urination sound from a smartphone or a high-sensitivity microphone into a water container; preprocessing reduces background noise and scales the signal; and feature extraction derives key features such as spectral features, Fast Fourier Transform (FFT), and MFCC for use as inputs;

[0029] (ii) a pattern classification / parameter estimation step including quantitative and qualitative evaluation of the flow pattern, wherein feature data are provided to the signal processing and intelligent model unit to estimate uroflowmetry parameters comprising Qmax, Qave, and Voided Volume, and, where defined, abnormality classification is performed using deep learning on audio features to indicate normal / abnormal patterns;

[0030] (iii) a testing or similarity assessment step to verify consistency with stored reference profiles / thresholds by comparing parameters and / or feature vectors with reference profiles / thresholds stored in a database, including tabular data (Excel / CSV / MySQL / PostgreSQL), user records, audio files, and predictions, to evaluate similarity or abnormality; and(iv) a decision-making and result presentation step including summarization, graph refinement, and publishing / storage, wherein post-processing smooths the flow-rate graph and formats outputs, results are displayed via the user interface and reporting module, and results are uploaded / stored via the database / IoT / IoMT integration module to a cloud or central database to enable real-time follow-up by specialists prior to termination of operation.

[0031] To address the foregoing, a system and a method for analyzing an audio signal to evaluate a urinary flow rate (FLOWMIND-RA) are developed such that a user can use a smartphone or a general audio recording device to record urination sound, and the system automatically and wirelessly processes the sound to calculate uroflowmetry parameters, such as maximum flow rate (Qmax), average flow rate (Qave), and voided volume, in a timely manner, thereby enabling screening and monitoring of urinary tract abnormalities anytime and anywhere, reducing travel burden and increasing follow-up frequency.

[0032] In the invention, the user records sound via a smartphone or a general recording device, and an audio input module forwards the data to a signal processing and Al engine to perform preprocessing and feature extraction (e.g., MFCC and Zero Crossing Rate (ZCR)) prior to estimating uroflowmetry parameters including Qmax, Qave, and voided volume. The results undergo post-processing and are displayed in a user interface and reporting module (13), and are connected to and stored through a database / IoT / IoMT integration module (12). The system structure is divided into a frontend layer, a backend and business logic layer (31), a machine learning pipeline, and a database layer, and the workflow corresponds to the flowcharts (20)-(27), aiming to enable convenient and reliable self-screening and continuous monitoring outside healthcare facilities.

[0033] The processing workflow includes: (i) component and key-feature analysis including signal acquisition, preprocessing, and feature extraction; (ii) pattern classification and parameter estimation via the signal processing and Al engine to estimate Qmax, Qave, and voided volume and, where defined, perform abnormality classification; (iii) testing or similarity assessment against reference profiles / thresholds stored in the database; and (iv) decisionmaking and result presentation including post-processing, graph refinement, display, and secure upload / storage for real-time follow-up.

[0034] Advantageous Effects of Invention

[0035] The invention enables uroflowmetry-like evaluation without requiring a dedicated in-hospital device, improves convenience, supports more frequent monitoring in real-lifeenvironments, and allows secure centralized data linkage for timely follow-up by specialists, thereby supporting preventive healthcare and digital health / telemedicine use cases.

[0036] Brief Description of Drawings

[0037] Fig.l An overview of the system and the method for analyzing an audio signal to evaluate a urinary flow rate.

[0038] Fig.2 A workflow flowchart of the system and the method for analyzing an audio signal to evaluate a urinary flow rate.

[0039] Fig.3 Components of the system and the method for analyzing an audio signal to evaluate a urinary flow rate.

[0040] Fig.4 Components of the frontend of the system and the method for analyzing an audio signal to evaluate a urinary flow rate.

[0041] Fig.5 Components of the backend of the system and the method for analyzing an audio signal to evaluate a urinary flow rate.

[0042] Fig.6 Steps of audio analysis of the system and the method for analyzing an audio signal to evaluate a urinary flow rate according to the present invention, showing a workflow of the machine learning pipeline.

[0043] Fig.7 Components of the database layer of the system and the method for analyzing an audio signal to evaluate a urinary flow rate.

[0044] Description of Embodiments

[0045] The following description of the invention is provided as one example with reference to the accompanying drawings to more clearly understand the operation of the system and the method. Identical parts or components in these drawings are denoted by the same reference numerals. This description does not limit the scope of the invention in any manner, and the scope of the invention is as defined in the appended claims.

[0046] Referring to Fig. 1, an overview of the system and the method for analyzing an audio signal to evaluate a urinary flow rate comprises:

[0047] 1. An audio input system (10): receives and records sound during urination from a suitable recording device into a collection container, while maintaining signal integrity for subsequent processing;2. A signal processing and intelligent model unit (11): performs noise reduction, signal preparation, feature extraction, and calculation of urinary flow parameters including maximum flow rate, average flow rate, and voided volume, for evaluating urinary tract conditions;

[0048] 3. A database and external device connectivity module (12): transmits and stores resulting data to a central database or a cloud system, and connects to related external devices or services, while governing security and access rights as specified; and

[0049] 4. A user interface and reporting unit (13): displays a flow- rate-versus-time graph and summarizes parameters for interpretation, prepares reports for users and medical personnel, and supports forwarding of results through specified channels.

[0050] Referring to Fig. 2, a workflow flowchart of the system and the method for analyzing an audio signal to evaluate a urinary flow rate comprises:

[0051] 1. Start of system operation (20): establishes a work session, checks readiness of the frontend (30), the backend (31), and the database layer (33), and sets parameters for audio data preparation (60) and feature extraction (61);

[0052] 2. User uploads a urination audio file (21): uploads via a file receiving unit (41) into the audio input system (10), checks file type and size, records metadata at an initial audio file processing unit (52), and stores related entries in the database layer (70)— (72);

[0053] 3. Preprocessing (22): reduces noise, performs high-frequency emphasis prior to analysis, scales signal level, and computes duration within the audio data preparation stage (60) under the signal processing and intelligent model unit (11);

[0054] 4. Audio feature extraction (23): computes zero crossing rate, mel-scale cepstral coefficients, and groups of spectral features, chroma, and tonnetz, and forms a feature vector according to the feature extraction stage (61);

[0055] 5. Parameter estimation using a model and abnormality classification (24): scales features using a trained scaler, performs inference according to the model processing stage (62) to obtain maximum flow rate, average flow rate, voided volume, and other related values, and stores the results in predictions (73);

[0056] 6. Post-processing and flow-rate graph generation (25): anchors graph scaling by the computed maximum flow rate, smooths the curve and adjusts tail portions according to the postcalculation processing stage (63), displays results at the user interface and reporting unit (13), and prepares connect! vity / storage via the database and external device connectivity module (12), with recording in (70)— (73);7. Display and reporting, or connection to the Internet of Things / Internet of Medical Things (26): forwards analysis results to an external system via the module (12) and stores a copy in the database layer (33); and

[0057] 8. End of operation (27): closes the work session, summarizes status, and records related events according to requirements.

[0058] Referring to Fig. 3, components of the system and the method for analyzing an audio signal to evaluate a urinary flow rate comprise:

[0059] 1. A frontend (30) including an audio file receiving unit (41), a parameter and flow-rate graph display unit (42), and external service connection units (43)(44), which receive an audio file from a user, transform a request and send it to the backend (31), and receive processed results for display as a flow-rate-versus-time graph and related parameters at the user interface and reporting unit (13);

[0060] 2. A backend and business logic layer (31) including a command or file receiving unit (51), an initial audio file processing unit (52), an interface unit to the artificial intelligence processing unit (53), an event and error logging unit (54), and a request exchange service unit (55), which verify audio file validity, control workflow ordering, coordinate operation among the frontend (30), the artificial intelligence processing unit (32), and the database layer (33), and record required events;

[0061] 3. An artificial intelligence processing unit (32) including an audio signal preparation stage (60), a feature extraction stage (61), a model processing stage (62), and a post-calculation processing stage (63), wherein:

[0062] • the audio signal preparation stage (60) prepares the audio signal by noise reduction, high-frequency emphasis prior to feature extraction, signal level scaling, and total audio duration calculation;

[0063] • the feature extraction stage (61) extracts features including zero crossing rate, mel- scale cepstral coefficients of 20-30 dimensions, groups of spectral features, chroma, and tonnetz, and forms a feature vector;

[0064] • the model processing stage (62) scales the feature vector using a trained scaler and calculates urinary flow parameters including maximum flow rate, average flow rate, and voided volume, together with related values as specified, and may classify flow patterns as appropriate; and• the post-calculation processing stage (63) performs post-calculation processing by smoothing the graph and adjusting tail portions, anchoring graph scaling by the computed maximum flow rate, and then sending results to the user interface and reporting unit (13) and storing the results in the database layer (33);

[0065] 4. A database layer (33) including a tabular / relational data structure (70), user records (71), audio files (72), and predictions (73), which receive and transmit data with the backend (31), store required results and metadata, support retrospective review and verification, provide data to the user interface and reporting unit (13), and support connection via the external module (12) according to the workflow (20)— (27).

[0066] Referring to Fig. 4, components of

[0067] the frontend (30) comprise:

[0068] 1. The audio file receiving unit (41): receives an audio file from a user via a display interface and checks completeness of information before sending a request to the backend (31);

[0069] 2. The display unit (42): displays a flow-rate-versus-time graph and uroflowmetry parameters and summarizes results from processing for interpretation by users and medical personnel, in conjunction with the user interface and reporting unit (13); 3. The external connection unit (43): forwards results or summaries to specified external services and communicates with related systems or devices through authorized connections; and

[0070] 4. Frontend technology (44): uses suitable development frameworks and tools to implement the user interface and support interoperability with the backend (31) and the database layer (33) according to the workflow (20)— (27).

[0071] Overall functions of the frontend (30) include receiving user data, creating requests to the backend (31), receiving processed results, presenting results via the display unit (42), and supporting result forwarding via the external connection unit (43), wherein communication and presentation comply with access control and data protection measures as specified.

[0072] Referring to Fig. 5, components of

[0073] the backend and business logic layer (31) include sub-units coordinated with the frontend (30), the artificial intelligence processing unit (32), and the database layer (33) as follows:1. The command receiving unit (51): receives a request and / or an audio file from the user via the frontend (30), checks authorization, assigns a work reference code, queues requests, and forwards the request to a subsequent stage according to the workflow (20)-(27);

[0074] 2. The initial audio file processing unit (52): checks file type, file size, signal length, and file header, standardizes format, sets sampling rate, checks signal level, masks personal data, and records metadata in the database layer (33) in entries (70)-(72); 3. The machine learning model interface unit (53): coordinates operation with the artificial intelligence processing unit (32), calls processes (60)-(63), inputs the prepared signal or feature vector into the model, receives calculated parameters including maximum flow rate, average flow rate, voided volume, and classification results, forwards the results to the user interface and reporting unit (13), and stores the results in predictions (73);

[0075] 4. The event / error logging unit (54): records events by time, related users, status at each stage (20)-(27), error items, rollback measures, notifications, and supporting evidence to enable retrospective auditing; and

[0076] 5. The request exchange service unit (55): specifies technologies for providing services without limitation, including Python, Flask, or FastAPI, to enable request exchange with the frontend (30) and orderly communication with the processing unit (32) and the database layer (33).

[0077] Overall functions of the backend (31) include controlling workflow, receiving and validating audio files, preparing data, forwarding tasks to the artificial intelligence processing unit (32), receiving results and formatting results for display at (13), and storing results and connecting to the external module (12), under supervision of security, data protection, and event logging requirements.

[0078] Referring to Fig. 6, steps of audio analysis of the system and the method for analyzing an audio signal to evaluate a urinary flow rate according to the present invention, showing a workflow of the machine learning pipeline, comprise:

[0079] 1. Audio data preparation (60), including operations prior to feature extraction as follows:

[0080] 1.1 Noise reduction to suppress background noise and interference unrelated to urination according to a defined method;1.2 Pre-emphasis using a defined audio library process to emphasize high-frequency bands and reduce signal distortion prior to feature extraction;

[0081] 1.3 Scaling energy level of the signal to an appropriate range for computation; and 1.4 Calculating total audio duration of the audio file for use in result interpretation. 2. Feature extraction (61), extracting quantitative representations of the audio signal to provide to the model as follows:

[0082] 2.1 Zero crossing rate to indicate frequency of sign changes from positive to negative; 2.2 Spectral features including spectral centroid, spectral bandwidth, roll-off, and spectral contrast to reflect frequency-domain characteristics;

[0083] 2.3 Mel-scale cepstral coefficients of multiple dimensions with temporal aggregation such as mean and standard deviation over time; and

[0084] 2.4 Chroma and tonnetz to analyze signal structure from a harmonic perspective. 3. Model processing / inference (62), transforming the feature vector into urinary flow parameters as follows:

[0085] 3.1 Data shaping to create a data table from computed features so that a feature list matches model requirements;

[0086] 3.2 Feature scaling using a pre-trained scaler to match training conditions of the model; 3.3 Invoking a trained model to compute inference results; and

[0087] 3.4 The predicted outputs comprising:

[0088] • maximum flow rate (milliliters per second);

[0089] • average flow rate (milliliters per second);

[0090] • total urine volume (milliliters);

[0091] • actual flow time (seconds);

[0092] • time to maximum flow rate (seconds);

[0093] • flow rate during the first two seconds (milliliters per second); and

[0094] • flow acceleration (milliliters per second squared).

[0095] Note: the above wording describes appropriate practice for persons skilled in the art and does not limit the scope of the invention to any particular tool, method, or parameter.

[0096] 4. Post-calculation processing (63), transforming numerical results for display and storage as follows:4.1 Anchoring the graph with predicted values by using the computed maximum flow rate as a reference, and scaling the square-root line of the mean to obtain a flow-rate-versus-time graph;

[0097] 4.2 Smoothing the graph and adjusting tail portions so that the flow pattern corresponds to an actual flow condition; and

[0098] 4.3 Delivering results to display by sending the graph and parameter values to the user interface and reporting unit (13) according to (25), and storing or forwarding the results via the database and external device connectivity module (12) according to (26) prior to proceeding to end of operation (27).

[0099] Referring to Fig. 7, components of

[0100] the database layer (33) comprise:

[0101] 1. A tabular / relational database file (70): stores numeric data and metadata in tabular formats including Excel, CSV, MySQL, and PostgreSQL, with index structures supporting search and retrospective verification;

[0102] 2. User records (71): store user codes, date and time, reference information, and processing status to link with audio files and predictions;

[0103] 3. Audio files (72): store uploaded source audio files with storage location, file size, file type, and signal length;

[0104] 4. Predictions (73): store parameter values and classification results from inference, together with model version and scaler, to support comparison and re-evaluation; and Data management rules of the database layer (33): define access rights and data protection according to requirements, record events related to creation, modification, and access, perform backups according to schedule, and connect for data exchange with the backend (31), the user interface and reporting unit (13), and the database / IoT / IoMT integration module (12) according to the workflow (20)— (27).

[0105] Operation process and practice comprise:

[0106] 1. Access and upload: the user accesses the frontend (30) via a web interface according to specified technology (44) and uploads an audio file via the audio file receiving unit (41) according to (20)— (21 ), wherein supported audio file types are accepted and file type and format are checked prior to forwarding;2. Signal preparation: the backend (31) receives the request at (51), verifies the file at (52), and forwards the file to the audio data preparation stage (60) to perform noise reduction, preemphasis, and signal level conditioning according to (22);

[0107] 3. Feature extraction: performed at (61) / (23) to compute zero crossing rate, groups of spectral features, multi-dimensional mel-scale cepstral coefficients, and chroma and tonnetz, and to generate a feature vector for inference;

[0108] 4. Scaling and inference: the trained scaler is used with the model in (62) / (24) to predict parameters including maximum flow rate, average flow rate, voided volume, actual flow time, time to maximum flow rate, flow rate during the first two seconds, and flow acceleration; 5. Abnormality classification: performed in (62) to identify normal or abnormal flow patterns based on learned pattern bases;

[0109] 6. Post-processing: the flow-rate-versus-time curve is adjusted in (63) / (25) by smoothing and tail adjustment;

[0110] 7. Graph generation and display: the graph and parameter values are sent for display at the user interface and reporting unit (13), with key parameters including maximum flow rate, average flow rate, voiding duration, and voided volume;

[0111] 8. Storage and connectivity: results are transmitted via the database / IoT / IoMT integration module (12) / (26) to be recorded in the database layer (33) in entries (70)— (73), supporting retrospective verification, re-analysis, and model improvement; and

[0112] 9. Summary to the user: the user views a summary table at the frontend (30) and has options to download or forward data to medical personnel via a LINE chatbot (43) prior to termination (27).

[0113] Reference Signs List

[0114] (10) Audio input system;

[0115] (11) Signal processing and intelligent model unit;

[0116] (12) Database / extemal connectivity module;

[0117] (13) Display and reporting unit;

[0118] (20)— (27) Workflow steps;

[0119] (30) Frontend;

[0120] (31) Backend and business logic;

[0121] (32) Machine learning processing;

[0122] (33) Database layer;(41) File intake unit;

[0123] (51) Command intake unit;

[0124] (52) Preliminary audio file processing unit; (53) Machine learning model interface unit; (54) Event / error logging unit;

[0125] (55) Service unit for request handling;

[0126] (60) Audio preparation;

[0127] (61) Feature extraction;

[0128] (62) Model inference;

[0129] (63) Post-calculation processing;

[0130] (70) Tabular / relational data store;

[0131] (71) User records;

[0132] (72) Audio files;

[0133] (73) Predictions.

Claims

CLAIMS1. A system for analyzing an audio signal to evaluate a urinary flow rate, comprising:an audio signal receiving unit configured to receive and record sound during urination from a recording device;a signal processing and intelligent model unit configured to process audio signal data received from the audio signal receiving unit to evaluate the urinary flow rate;a display and reporting unit configured to display a result of evaluating the urinary flow rate received from the signal processing and intelligent model unit in a form of urinary flow parameters versus time; anda database and external device connectivity module configured to transmit and store result data of evaluating the urinary flow rate to a central database or a cloud system, and to connect to an external device or an external service, wherein the signal processing and intelligent model unit is configured to process the audio signal data to evaluate the urinary flow rate by:performing preprocessing in which the audio signal received from the audio signal receiving unit is subjected to noise reduction and signal level scaling;performing audio feature extraction in which a signal after the preprocessing is used to calculate feature values selected from one or more of a zero crossing rate (ZCR), spectral features, mel -frequency cepstral coefficients (MFCC), a chroma feature, and a Tonnetz feature; andevaluating the urinary flow rate by inputting the calculated feature values to, a machine learning model to estimate the urinary flow rate in a form of urinary flow parameters.

2. The system of claim 1, wherein a nocturia parameter extraction module automatically:identifies nocturnal time windows;calculates nocturnal urine volume;determines a nocturnal polyuria index; andgenerates a nocturia severity classification.

3. The system of claim 1, further comprising a longitudinal analytics module configured todetect changes in voiding behavior trends over time.

4. The system of claim 1, wherein the urinary flow parameters are selected from one or more ofa maximum flow rate, an average flow rate, a voiding duration, a voided volume, a flow ratewithin a specified time interval, and a flow acceleration.

5. The system of claim 1 or claim 4, wherein the machine learning model is selected from an artificial neural network and deep learning.

6. A method for analyzing an audio signal to evaluate a urinary flow rate, comprising:receiving an audio signal by receiving sound during urination from a device and recording the sound by an audio signal receiving unit;processing the audio signal by processing audio signal data received from the audio signal receiving unit to evaluate the urinary flow rate by a signal processing and intelligent model unit; anddisplaying and reporting by displaying a result of evaluating the urinary flow rate obtained from the signal processing and intelligent model unit through a display and reporting unit in a form of urinary flow parameters versus time,wherein processing the audio signal data to evaluate the urinary flow rate comprises:preprocessing by performing noise reduction and signal level scaling of the audio signal received from the audio signal receiving unit;audio feature extraction by calculating feature values from a signal after the preprocessing,wherein the feature values are selected from one or more of a zero crossing rate (ZCR), spectral features, mel-frequency cepstral coefficients (MFCC), a chroma feature, and a Tonnetz feature; and evaluating the urinary flow rate by inputting the calculated feature values to a machine learning model to estimate the urinary flow rate in a form of urinary flow parameters.

7. The method of claim 6, wherein the urinary flow parameters are selected from one or more ofa maximum flow rate, an average flow rate, a voiding duration, a voided volume, a flow ratewithin a specified time interval, and a flow acceleration.

8. The method of claim 6 or claim 7, wherein the machine learning model is selected from an artificial neural network and deep learning.

Images