Multi-modal context-aware environmental health monitoring and recommendation system and method

Abstract

The present disclosure relates to a system and method for monitoring air quality and providing recommendation for managing air quality. The system comprises one or more environment monitors, a user device and a server in communication with the environment monitor. The system enables the user to input a contextual information to specify a space and / or use of space in which the environment monitor is placed. The system is configured to provide recommendation based on the designated contextual information. The system offers flexibility in selecting sensors and provide specific, actionable advice tailored to individual environments. The system integrates seamlessly with other smart home devices for automated air quality management. Furthermore, the system allows users to select sensors based on the materials they work with, the processes they apply, and the pollutants of concern within their workspace.

Description

MULTI-MODAL CONTEXT-AWARE ENVIRONMENTAL HEALTH MONITORING AND RECOMMENDATION SYSTEM AND METHODCROSS REFERENCE TO RELATED APPLICATIONThis application claims the priority benefit of the U.S. Patent Application No. 63 / 658,010 filed on June 10th, 2024, the contents of which are hereby incorporated by reference in their entirety.TECHNICAL FIELD

[0001] The present invention generally relates to a system for monitoring air pollution. More specifically, the present invention relates to a system and method for monitoring air pollution and providing air pollution risk advisory.BACKGROUND

[0002] Environmental monitoring involves identifying and analyzing environmental conditions to assess the impact of various activities on the environment using specialized devices. The current devices available for environmental monitoring have fixed sensing capacities and detect certain pollutants in a static manner. These devices do not consider the specific characteristics or conditions of the environment where they are placed. Historically, air quality monitoring devices were primarily designed for industrial and regulatory purposes, making them large, expensive, and requiring professional installation and maintenance.

[0003] Air quality monitoring devices typically detect pollutants based on levels of particulate matter and volatile organic compounds (VOCs). As a result, these devices might not distinguish between routine and hazardous environmental conditions.

[0004] Additionally, current air quality monitoring devices come with a predetermined set of sensors, which might not detect all relevant pollutants. Different work or creative spaces, such as wood shops, paint shops, and ceramic workshops, have unique pollutant sources, health risks, and preventive measures. Despite the high risk of harmful pollutants, including carcinogens, in creative spaces, these environments have been largely unaddressed by existing solutions. Integrated solutions like air purifiers with sensors often lack customization, leading to generalized recommendations that might not be suitable for all environmental conditions. Devices with narrow-range sensors that focus on particulate matter (PM) like PM2.5 and PM 10 may detect particles affecting human health but fail to identify and mitigate other harmful pollutants such as volatile organic compounds (VOCs), Nitrogen Oxides (NOx), Formaldehyde, carbon monoxide (CO) and carbon dioxide (CO2) or harmful settings such as high level of noise.

[0005] Therefore, there is a need for a system and method for environmental monitoring that is adaptable, customizable, and provide more targeted solutions for improving air quality.SUMMARY

[0006] In a first aspect of the present invention discloses a system and method for monitoring air quality and providing recommendation(s) for managing air quality. The system comprises one or more environment monitors, a user device, and a server in communication with the environment monitor. The user device is also referred as user input interface throughout this document. The user device is associated with a user. The system further comprises a database in communication with the server.

[0007] The environment monitor comprises an outer compartment, an encoder and push button, a switch, a screen protector and e-ink display and one or more ventilation holes. In some examples the outer compartment comprises a machined extruded aluminum, providing a robust and durable housing for the environmental monitor. The outer compartment could be made from plastic or other material. The ventilation holes enable airflow to the sensing unit, ensuring accurate and responsive environmental measurements. The e-ink display is equipped with a screen protector to safeguard the display from damage in workshop environments. The charging dock and integrated battery simplify power management and enhance portability. The encoder with the click button enables the users to navigate through various options. Further, the environment monitor comprises a screw assembly. The screws holding an inner compartment from sliding. Further, the screw assembly facilitates simple assembly and disassembly, allowing easy opening of the environment monitor for repair and maintenance.

[0008] The environment monitor further comprises a magnet, a battery, an inertial measurement unit, and an expansion port. The magnet is incorporated into the environment monitor to enable secure attachment of the environmental monitor to metallic surfaces in various environments. Thebattery is a power source of the environment monitor. The battery could be a rechargeable battery. The rechargeable battery could be a lithium-ion battery. The Inertial measurement unit enables detecting handling of the device, re-orienting for changing display orientation and modes of operation. Further, the expansion port comprises pogo pins. The pogo pins enables power and data transfer for additional modules or alternative modes of powering.

[0009] The environment monitor further comprises one or more sensor. The sensor, includes, but not limited to, a volatile organic compound sensor, a particulate matter sensor, a carbon dioxide sensor, a temperature sensor, a relative humidity sensor, a formaldehyde sensor, carbon monoxide sensor, a nitrogen oxides sensor, an ozone sensor, a radon sensor, a radiation sensor, an infrared thermal sensor, a passive infrared sensor, an ultraviolet light sensor, a sound sensor and an inertial measurement unit. Further, additional sensors can be attached and / or sensors can be replaced depending on the contextual information. The contextual data informs users regarding additional or relevant sensors to install and / or replace. The environment monitor further has the ability to expand on the visibility of the status of the environment data, through a display, comprising of, but not limited to, a large screen, and a physical display, such as a servo motor moving an indicator up and down. The environment monitor further comprises a wireless communication circuitry, a memory to store machine instructions and a processor.

[0010] The system further comprises a user interface to input contextual information about the space in which the environment monitor is placed. The user interface provides informational advice to the user about risk(s) associated with this particular space based on the contextual information. The space could be in a home, a manufacturing facility, a workshop space or a creative space including a wood shop, a recording studio, a fabric workshop, a metal fabrication shop, a jewelry workshop, an office space, a photography development room, a paint shop or a ceramics workshop space.

[0011] Further, the user interface prompts a user to change features and content displayed on a machine interface comprising of at least one of a mobile device, computer, and interface on the environment monitor, based on the contextual information. The system further comprises a processor-readable instructions. The processor-readable instructions comprising at least one composition of a mobile device, a natural language interface, a display on the environment monitorand a computer, an opportunity for a user to input a contextual information to specify a space and / or use of space in which the environment monitor is placed via machine interface. Further, the user interface prompts a user to change features and content presented through one or more machine interfaces based on the contextual information. The machine interfaces comprise any processor- enabled device or system capable of receiving, processing, and communicating information to users through any sensory modality and / or accepting user input. Machine interfaces include, but are not limited to, traditional computing devices such as mobile devices, tablets, desktop computers, and laptops; embedded interfaces such as displays integrated into the environment monitor; voice- activated systems including smart speakers, digital assistants, and conversational Al systems that provide purely audio-based interaction; immersive interfaces such as augmented reality (AR) devices, virtual reality (VR) headsets, smart glasses, and mixed reality systems; Internet of Things (loT) devices with user interaction capabilities; automotive interfaces; wearable devices including smartwatches, fitness trackers, and audio-only devices such as wireless earbuds or bone conduction headphones; ambient computing systems that provide feedback through environmental changes such as lighting, temperature, or sound; and emerging interface technologies that utilize natural language processing, gesture recognition, eye tracking, brain-computer interfaces, haptic feedback, olfactory output, or other sensory input / output modalities. The machine interfaces may operate independently or in combination to provide multimodal interaction experiences that can be entirely audio-based, entirely visual, entirely haptic, or any combination of sensory modalities.

[0012] The machine interfaces are characterized by their ability to execute processor-readable instructions and facilitate bidirectional communication between users and the environmental monitoring system through any combination of sensory modalities. Each machine interface comprises at least a processing unit capable of interpreting instructions, an input mechanism for receiving user commands or data (which may include but is not limited to touch, voice, gesture, eye movement, neural signals, or proximity detection), and an output mechanism for presenting information to users through any sensory channel. Output mechanisms may include but are not limited to visual displays, audio output (including speech synthesis, musical tones, sound effects, or ambient audio), haptic feedback (including vibration, force feedback, or tactile stimulation), olfactory output, temperature changes, or environmental modifications such as lighting adjustments. Notably, machine interfaces may operate entirely without visual components, providing complete functionality through audio-based interaction, haptic feedback, or other non-visual modalities. Themachine interfaces may utilize various communication protocols including but not limited to WiFi, Bluetooth, cellular networks, mesh networks, or direct wired connections to exchange data with the environment monitor and associated systems.

[0013] The system further comprises processor-readable instructions. The processor-readable instructions are designed to operate across diverse machine interface architectures and may be distributed across multiple processing units within the system. The processor-readable instructions comprise adaptive interface components that automatically configure themselves based on the capabilities and characteristics of available machine interfaces, enabling seamless operation whether the user interacts through a mobile device, a natural language interface, a display on the environment monitor, a desktop computer, an augmented reality headset, a voice-only assistant, audio-based wearables, ambient sound systems, or any combination thereof. This adaptive capability ensures that users can access contextual information input functionality and receive recommendations through their preferred or available interface modality, including purely audiobased interactions, visual-only interfaces, haptic-only feedback systems, or emerging technologies not specifically enumerated herein. The system automatically adapts its communication style, information density, and interaction patterns based on the sensory capabilities and limitations of the active machine interface(s).

[0014] The contextual information comprises at least one composition selected from a floor area of the space; a height of the space, a volume of the space; a type of product stored in the space, a type of tool in the space; a type of machine in the space; a type of activity typically the space is used for; a type of activity occurring in the space; a type of materials processed in the space; a type of furniture in the space; a type of ventilation system available in the space; a type of air filtration system available in the space; a geographic location of the space; a type of building the space is in; and a type of heating and cooling system available in the space. The space could be a workshop space or a creative space including a wood shop, a paint shop or ceramics work shop space.

[0015] The user input format for the contextual information includes, but not limited to, an audio, an image, a video, a text, a GPS location, a wireless signals, a Bluetooth, a Wi-Fi status, a data from lidar sensor, and a data from depth sensing sensor. The user input could be an image of the space captured by a camera. The user input could be through a voice audio file describing the usage of thespace and the machines present. The user input via a list of options available on a mobile software application, the contextual information extracted, via a machine learning algorithm, features from an input file. The machine learning algorithm capable of processing different formats of input to detect features and output characteristics of the space, and tailored to specify a space and / or use of space in which the environment monitor is placed. The characteristics of the space includes, but not limited to, usage of the space, tools and machines present in the space, approximate volumetric dimensions of the space or materials in space.

[0016] The user input format for the contextual information includes advanced input modalities enabled by modern machine interfaces. These advanced modalities include voice commands processed through natural language understanding, gesture -based inputs captured through computer vision or motion sensors, eye tracking data from gaze-enabled interfaces, spatial mapping data from augmented or virtual reality systems, biometric inputs including heart rate or stress indicators from wearable devices, environmental sensing data automatically captured by smart home systems, location and movement patterns from GPS and accelerometer data, and contextual data inferred through machine learning analysis of user behaviour patterns across multiple interface touchpoints. The system's machine learning algorithms are trained to interpret and extract meaningful contextual information from any of these input modalities, either individually or in combination, providing a comprehensive understanding of the user's environment and needs.

[0017] The processor-readable instructions further receive a recommendation to adjust alert profiles associated with the contextual information via at least one of the machine interfaces. Further, the processor-readable instructions provide the user input to the machine interface and accept the recommendation to adjust alert profiles associated with the contextual information. Further, the processor- readable instructions send an instruction to set the adjusted alert profile associated with the contextual information based on the acceptance of the recommendation received via the machine interface of the environmental monitor, mobile and / or computer interface.

[0018] Further, the processor-readable instructions store and / or retrieve data to / from a database. Further processor-readable instructions store and / or retrieve data comprising at least one composition selected from the environment monitoring sensor, a processor-generated timestamp, and the contextual information. The processor-readable instructions further based on contextualinformation determine and analyze live and historical data collected from the environment monitor, and / or from user input (comprises but not limited to a question about the data from the environmental monitor, the health risks, and / or actions they can take in and around the specific context input through a natural language interface). Further, the processor-readable instructions analyzing a user profile for the user and determining one or more categories from the user profile that are related to the contextual information and / or data stored in the database via the processor. Categories can comprise of, but not limited to, the type of profession, type of activity, type of materials handled, demographic profile, health risk profile, a level of health risk (for example a value between 1-10). Further, the processor-readable instructions provide the recommendations to mitigate environmental health risks associated with the contextual information. Further, one feature example could be changing alert threshold settings for the user to be notified accordingly to health and safety recommendations related to pollutants likely to be present in the space designated, based on the contextual information. Further, another feature is an alert and recommendation, on measures to reduce health risks and / or integration and optimisation of automated solutions, presented to users when a pollution profile exceeds advised levels.

[0019] The above summary contains simplifications, generalizations and omissions of detail and is not intended as a comprehensive description of the claimed subject matter but, rather, is intended to provide a brief overview of some of the functionality associated therewith. Other systems, methods, functionality, features and advantages of the claimed subject matter will be or will become apparent to one with skill in the art upon examination of the following figures and detailed written description.BRIEF DESCRIPTION OF THE DRAWINGS

[0020] FIG. 1 exemplarily illustrates an environment of a system for monitoring air quality and providing recommendation for managing air quality, according to some examples of the present invention.

[0021] FIG. 2 exemplarily illustrates a block diagram of the system for monitoring air quality and providing recommendation for managing air quality., according to some examples of the present invention.

[0022] FIG. 3 exemplarily illustrates a block diagram of an environment monitor, according to some examples of the present invention.

[0023] FIG. 4 exemplarily illustrates a perspective view of the environment monitor, according to some examples of the present invention.

[0024] FIG. 5 exemplarily illustrates another perspective view of the environment monitor, according to some examples of the present invention.

[0025] FIG. 6 exemplarily illustrates another perspective view of the environment monitor, according to some examples of the present invention.

[0026] FIG. 7 exemplarily illustrates an assembly of screws of the environment monitor, according to some examples of the present invention.

[0027] FIG. 8 exemplarily illustrates an inner compartment sliding into an outer compartment of the environment monitor, according to some examples of the present invention.

[0028] FIG. 9 exemplarily illustrates components of the environment monitor disposed at the inner compartment, according to some examples of the present invention.

[0029] FIG. 10 exemplarily illustrates a perspective view of the environment monitor with an expansion for formaldehyde sensor, according to some examples of the present invention.

[0030] FIG. 11 exemplarily illustrates an exploded view of the environment monitor with an expansion for formaldehyde sensor, according to some examples of the present invention.

[0031] FIG. 12 exemplarily illustrates the environment monitor mounted onto a metallic machine via magnet, according to some examples of the present invention.

[0032] FIG. 13 exemplarily illustrates the display of the environment monitor displaying various data regarding the environment, according to some examples of the present invention.

[0033] FIG. 14 exemplarily illustrates a flowchart of a method for monitoring air quality and providing recommendation for managing air quality, according to some examples of the present invention.

[0034] FIG. 15 exemplarily illustrates the flowchart of a method for providing recommendation for managing air quality, according to some examples of the present invention.

[0035] FIG. 16 exemplarily illustrates a screenshot of a user interface to enable a user to select information regarding the regarding the environment of a workspace or creative space, according to some examples of the present invention.

[0036] FIG. 17 exemplarily illustrates a screenshot of a user interface displaying a real time data of the workspace and a button for dynamic data labelling, according to some examples of the present invention.

[0037] FIG. 18 exemplarily illustrates a screenshot of a user interface displaying recommendations, according to some examples of the present invention.

[0038] FIG. 19 exemplarily illustrates a screenshot of a user interface displaying the details of the workspace or creative space, according to some examples of the present invention.

[0039] FIG. 20 exemplarily illustrates a screenshot of a user interface displaying analytics of an environment, according to some examples of the present invention.

[0040] FIG. 21 exemplarily illustrates a screenshot of a user interface displaying a real time data and prompts for action, according to some examples of the present invention.

[0041] FIG. 22 exemplarily illustrates a screenshot of a user interface displaying a context aware recommendation to the user, according to some examples of the present invention.

[0042] FIG. 23 exemplarily illustrates a screenshot of a user interface displaying the high PM2.5 levels indicating poor air quality, according to some examples of the present invention.

[0043] FIG. 24 exemplarily illustrates a screenshot of a user interface displaying low PM2.5 levels indicating good air quality, according to some examples of the present invention.

[0044] FIG. 25 exemplarily illustrates a screenshot of a user interface providing data visualization of particulate matter (PM) concentration related to specific activities in a ceramic studio, according to some examples of the present invention.DETAILED DESCRIPTION OF EXAMPLE EMBODIMENTS

[0045] Referring to FIG. 1 and FIG. 2, an environment 100 of a system comprises one or more environment monitors 102, a user device 104 and a server 106 in communication with the environment monitor 102. The user device 104 is also referred as user input interface 104 throughout this document. The user device 104 is associated with a user. The system further comprises a database 108 in communication with the server 106. In some examples, the environment monitor 102, the user device 104 and the database 108 are connected to the server 106 via a network 110.

[0046] The environment monitor 102 is an intelligent, network-connected, multi- sensing environmental monitor 102 configured to detect changes to health risks in a home or work environment. The environment monitor 102 comprises one or more environment sensors to sense environmental data. The environmental data includes, but not limited to, air quality, particulate matter, temperature, humidity, formaldehyde, volatile organic compounds, carbon monoxide, ozone, radon, radiation, carbon dioxide, ultraviolet light intensity, sound (dB level) and thermal heat. The user input interface is configured to store and analyze contextual information along with the timestamped, environmental data.

[0047] The user input interface 104 allows users to input contextual information into the environment monitor 102. The contextual information comprises at least one composition selectedfrom a floor area of the space 114, a height of the space 114, a volume of the space 114; a type of product stored in the space 114, a type of tool in the space 114, a type of machine in the space 114, a type of activity typically used in the space 114, a type of activity occurring in the space 114, a type of materials processed in the space 114, a type of furniture in the space 114, a type of ventilation system available in the space 114, a type of air filtration system available in the space 114, a geographic location of the space 114, a type of building present in the space 114, and a type of heating and cooling system available in the space 114. The system further configured to receive contextual information using the outdoor monitor 116, motion sensor 118, wireless network router 122, camera 124, microphone (in smart devices like Alexa) and electrical current sensor 120. The server 106 is configured to process the contextual information at the processor 128 and provide analytics 126. The server 106 is configured to provide recommendation based on the designated contextual information. In another embodiment, the computing device 112 is configured to process the contextual information and provide recommendation based on the designated contextual information.

[0048] The contextual information refers to data that provides additional background and specifics about the space 114 in which the environment monitor 102 is deployed. This information helps in understanding the unique characteristics and conditions of the space 114, which could influence the interpretation of sensor data and the generation of relevant recommendations. The contextual information can include physical characteristics such as space type, ventilation products installed, machines, materials and products used, processes and operations, location and pollutant sources. The space 114 refers to an environment requiring air pollution risk advisory.

[0049] The information related space type includes, but not limited to, data related to general use of area, for example, ceramics space, wood workshop, jewellery workspace, fabric space, welding area, home office, kiln room and size of the space 114, including floor area, height, and volume. The information related to ventilation includes, but not limited to, type and efficiency of the ventilation system in place, details about air filtration systems, including type and capacity, and equipment and tools. The information related to machines, includes, but not limited to, types of machinery present in the space 114, and specific tools, materials and activities used in the space 114 that might influence air quality. For example, the machinery includes lathe, kiln, and metal casting tools. The information related to process and operations, includes, but not limited to,common activities performed in the space 114, types of materials processed or used, and geographical and structural information. For example, the common activities include welding, soldering, mopping floors and firing a kiln. For example, the material being handled at the space 114 includes types of wood, metals and ceramics. For example, specifying whether the wood is a hardwood or a soft wood, or the species of the wood, enables an understanding of whether the dust generated from the specified wood is a carcinogen, asthmagen or not.

[0050] The information related to location, includes, but not limited to, geographic location of the space 114, which can impact environmental conditions, type of building in the, and health and safety data. For example, the type of building includes commercial and residential building. The information related to pollutant sources includes, but not limited to, potential sources of pollutants within the space 114, and exposure risks including information about potential health risks associated with activities and materials. The geographic location supports estimating risks of pollution entering the building from the outdoors, and is also capable of taking in forecast air pollution information to communicate outdoor pollution risks influencing the indoor environment.

[0051] In another embodiment, the system is configured to enable the user to take an image of the space 114 as a user input to the contextual information extraction system. The image is processed using an Al algorithm, with natural language capabilities to extract features and contextual information from the image. For example, a user could take a photo of a piece of wood being machined, the Al algorithm, can output the type of wood it is likely to be, such as a hard wood or soft wood. Then lets an algorithm output the health risks of the type of wood in the image. Another example is if a user inputs a photo of a space, such as a ceramics studio, the Al algorithm can detect a window in the space and different types of clay, kiln and tools in the space, enabling contextual data to be extracted into the environment monitoring system.

[0052] In some examples, the user device 104 comprises at least one of inertial measurement unit (IMU), camera 124 and depth sensors, for example, lidar. The user device 104 enables to map out the volumetric dimensions of the space 114. The data could be converted into a 3D spatial model.

[0053] In yet another embodiment, the system is configured to enable the user to enter contextual information through voice using the user device 104. The voice is processed using an Al algorithmwith speech recognition capability to convert the voice into annotated text, and key features about the space 114 could be identified within this text. In some examples, the system is configured to store the contextual information on the computing device 112 or a remote database 108. The contextual information further includes cumulative hours of operation of the environment monitor and / or machine the environment monitor is attached to in the space 114. The information related to the cumulative hours of operation in the space 114 will be processed and stored. This data can be used to assess operational lifetime of environment monitor and / or machine the environment monitor is attached to in order to prompt users to run routine maintenance, and replace components (for example the sensors) if nearing the end of life.

[0054] In some examples, the contextual information could be categorized into two types, including a static contextual information and a dynamic contextual information. This enables users to outline the context of the space 114 on first onboarding of installing the monitor 102, and describe the space characteristics. The dynamic contextual information enables users to input real-time or periodically changing information about activities and conditions within the space 114. This information allows the environment monitor 102 to adapt its monitoring and recommendations based on the current state of the space 114.

[0055] In some examples, the ongoing activities that may influence air quality or environmental conditions includes firing the kiln, which generates significant heat and may release particulates or fumes, impacting air quality; mopping the floor in a clay room, which could stir up dust and particulates, affecting respiratory health; wood turning produces wood dust, which could be hazardous if inhaled over time; metal casting involves high temperatures and potential release of metal fumes, soldering electronics generates fumes that may contain harmful substances like lead or rosin; varnishing releases volatile organic compounds (VOCs) that can affect air quality; gluing wood may emit fumes from adhesives, impacting indoor air quality; and photo development in a dark room uses chemicals that can release fumes or vapors. The factors influencing air quality further includes environmental conditions such as temperature fluctuations due to specific activities, humidity changes influenced by processes like drying or cooling, and ventilation adjustments based on activity intensity. The factors influencing air quality further includes equipment usage.

[0056] Further, in some examples, the users have the option to input contextual information into the environment monitoring system, the data on the operation of specific machines or tools in the space, which include run times for equipment like kilns, lathes, or casting tools and status of air filtration and ventilation systems. Further, the users have the option to input contextual information about the user’s temporary health conditions, including instances of sneezing, which might indicate the presence of allergens and occurrence of headaches, potentially signaling poor air quality or high levels of pollutants.

[0057] The dynamic contextual information is sent to the database 108 with a timestamp. This information can then be plotted on a graph of the environment monitor data to indicate when specific events occurred. This visualization helps users easily identify periods of high pollution and correlate them with space usage. Additionally, tagging this information facilitates the correlation of events, enabling more accurate data analysis and the training of machine learning algorithms to classify different types of events. The activities input can be facilitated on the user device 104 through prompting users from a list, based on the static contextual information. In some examples, a natural language interface, can prompt based on the contextual information.

[0058] The dynamic contextual information can be entered any time, before, during and / or after the event has occurred. In some examples the environment monitor 102 will process the dynamic contextual data and “tag” it with a timestamp, to provide tailored recommendations. For example, if the static contextual information of the space 114 is a ceramics space, and the dynamic contextual information is that the user is mopping the floor, and environmental monitoring data shows high levels of particulate matter, the recommendation and advice could inform users about the risks of silica dust in ceramics spaces, that can lead to silicosis, a permanent damage to the lungs, and users can wear a respirator mask N95, use a wet-mop, open windows and turn on all filtration systems. In another example in a wood workspace, if users are machining wood and pollution levels are increasing, and normally the filtration system performs well to maintain low pollution levels, the recommendation can advise users to change the filter in the filtration device or check that the device is functioning and installed properly without any gaps in the filter media.

[0059] In some examples, a method of sensing presence of a user in the space 114, via a presence sensor, thermal imaging sensor, camera 124, or by deciphering user GPS location, Wi-Fi / Bluetoothactivity of user’s phone nearby (or a combination of these sensing data) to collect store timestamps in the database 108 of when users were in the vicinity of the environmental monitor 102. This enables processing of data, to calculate either a user’s or space occupants’ average exposure to pollutants in the space. For example, recommend a time- weighted average (TWA) of 8 hours for a working day to be below a threshold of pollution exposure. The data can be presented to users in a simplified method of TWA of the day, enabling users to take proactive action to ensure low pollution exposure.

[0060] The addition of dynamic contextual information enables risk analysis through potential pollutants that may be present in the space. For example, if a furniture maker is assembling a chair with a glue, the system can prompt users to input what glue they’re using (through an image, text or voice) (image for example of the material contents, MSDS sheet) which can then be analysed to assess whether it contains formaldehyde, a carcinogen. The context information allows for understanding potential sources of pollutants to support data collected from environment monitors 102. In another example, the user can input the type of wood they are working with, and understand whether the wood dust from the type of wood is a carcinogen or an asthmagen.

[0061] Referring to FIG. 3 to 14, the environment monitor 102 comprises a processing unit 130, a memory 132, a clock 134 and a visual feedback interface 136. The environment monitor 102 further comprises a wireless communication interface 138, a power source 140, one or more environment monitoring sensors 142 and a display 180. The environment monitor 102 comprises an outer compartment 152 and an inner compartment 160. Further, the outer compartment 152 has a cavity and an opening at the top of the outer compartment 152. Referring to FIG. 8, the inner compartment 160 slides into the cavity through the opening of the outer compartment 152.

[0062] The environment monitor 102 further comprises a screen protector and e-ink display 144, an encoder 146 with a click button 148, one or more ventilation holes 150 for internal air quality sensor, and a switch 154. The screen protector 144 provides safeguard to an e-ink display 144 from damage in workshop space or creative space environments. The ventilation holes 150 airflow to the sensing unit. Further, the ventilation hole 150 ensures accurate and responsive environmental measurements. Referring to FIG. 7, the screws 158 hold the inner compartment 160 from sliding. Further, the screw 158 facilitates simple assembly and disassembly, allowing easy opening of theenvironment monitor 102 for repairs and maintenance. Further, the outer compartment 152 could be made of machined extruded aluminum that provides a robust and durable housing for the environment monitor 102.

[0063] Referring to FIG. 8 the environment monitor 102 comprises an inlet and outlet (166,164), a microprocessor 162, and printed circuit board (PCB) 172. The e-ink display 144 is configured to display custom information. The custom information includes, but not limited to, a clock, a local weather forecast, an outdoor temperature, an action recommendation, or an air quality data from the integrated sensors. Further, the e-ink display 144 could also be integrated with a calendar, which enables to display the current schedule and next upcoming schedule in the calendar.

[0064] The e-ink display 144, providing an energy-efficient method for displaying sensor information. The environment monitor 102 further comprises a combination of colour LEDs to display real-time air quality information through a color-coded traffic light system and visually alerts users to environmental risks in the space. In another embodiment, the environment monitor 102 further includes a speaker that provides auditory feedback to alert users exposed to pollution exceeding a set threshold.

[0065] The environment monitor 102 utilizes a Wi-Fi network chip, such as the ESP32, for wireless network communication to send data to and from a data storage device. The ESP32 is known for robust performance, low power consumption, and integrated support for both Wi-Fi and Bluetooth, making it an ideal choice for continuous environmental monitoring. In addition to Wi-Fi, the environment monitor 102 could employ other communication protocols to enhance flexibility and ensure reliable data transmission. For instance, Zigbee, a low-power, low-data-rate wireless network standard, could be used for its mesh networking capabilities, allowing for extended range and redundancy. Further, a unified IP-based connectivity protocol for smart home devices, could also be integrated to ensure interoperability across different brands and devices. LoRaWAN, known for its long-range, low-power communication, is suitable for wide-area networks where devices are spread over large distances. Additionally, cellular signals, such as 3G, could be utilized to provide connectivity in areas without Wi-Fi coverage, ensuring that the environment monitor 102 could transmit data even in remote or mobile settings. Each of these methods offers unique advantages depending on the specific deployment requirements and environmental conditions.

[0066] The environment monitor 102 further enables users to input contextual information into the system by navigating the display 144 using a user interface. In some examples, the user interface consists of an encoder 146 with a click button 148 functionality, which allows users to navigate through various options. In another embodiment, the system enables users to enter the contextual information through a mobile application. In the mobile application, users could select contextual information.

[0067] The environment monitor 102 further comprises one or more air quality related sensors 174. The sensors 174, include, but not limited to, a volatile organic compound sensor, a particulate matter sensor, a carbon dioxide sensor, a temperature sensor, a relative humidity sensor, a formaldehyde sensor, carbon monoxide sensor, a nitrogen oxide sensor, an ozone sensor, a radon sensor, a radiation sensor, infrared thermal sensor, passive infrared sensor, thermal / PIR sensor, auditory sensor (for example measuring sound levels in dB), GPS sensor and ultraviolet light sensor. The auditory sensor comprises a sensor configured to detect, measure, or respond to acoustic phenomena including sound waves, acoustic vibrations, pressure waves, or sound-related phenomena. The auditory sensor may operate in audible frequency ranges (approximately 20 Hz to 20 kHz), ultrasonic frequency ranges (above 20 kHz), or infrasonic frequency ranges (below 20 Hz). The auditory sensor includes, but is not limited to, sound level sensors configured to measure decibel levels, microphones (including condenser, dynamic, electret, or MEMS microphones), piezoelectric sensors that convert mechanical vibrations to electrical signals, acoustic pressure sensors that detect sound pressure variations, ultrasonic sensors capable of detecting frequencies above human hearing range, pressure wave sensors that measure acoustic pressure differences, acoustic emission sensors for detecting high-frequency stress waves, contact microphones for detecting vibrations through solid materials, accelerometers configured to detect vibrations correlating with sound, differential pressure sensors for measuring pressure wave variations, fiber optic acoustic sensors, or laser interferometric sensors for non-contact acoustic measurement.

[0068] The environment monitor 102 further comprises one or more air quality related sensors 174. The sensors 174, include, but not limited to, a volatile organic compound sensor, a particulate matter sensor, a carbon dioxide sensor, a temperature sensor, a relative humidity sensor, a formaldehyde sensor, carbon monoxide sensor, a nitrogen oxide sensor, an ozone sensor, a radon sensor, aradiation sensor, infrared thermal sensor, passive infrared sensor, thermal / PIR sensor, auditory sensor, GPS sensor and ultraviolet light sensor.

[0069] The auditory sensor is configured for multiple detection purposes including human presence detection and machinery operation classification. For human presence detection, the auditory sensor detects acoustic signatures in the human audible range to determine occupancy and calculate user exposure times. For machinery detection, the auditory sensor comprises capability to detect both audible frequency sounds (20 Hz to 20 kHz) and ultrasonic frequencies (above 20 kHz) to identify and classify specific equipment operations. The ultrasonic detection enables automated machinery identification while preserving user privacy by operating above human speech frequency ranges.

[0070] The auditory sensor data is processed using machine learning algorithms trained to classify specific machinery acoustic signatures including, but not limited to, lathe operations, drill operations, 3D printer operations, kiln firing cycles, air filtration system operation, ventilation fan operation, grinder operations, saw operations, and CNC machine operations. The machine learning classification algorithms may include neural networks, support vector machines, random forest classifiers, convolutional neural networks for frequency pattern recognition, or ensemble methods for improved accuracy. The acoustic classification system is trained on labeled datasets of machinery operation sounds and can adapt to new equipment through user-supervised learning.

[0071] In one embodiment, the system correlates detected machinery operations with real-time air quality measurements to automatically label pollution events and provide contextual recommendations. For example, detecting lathe operation sounds concurrent with increased particulate matter levels enables automated activity tagging and targeted safety advice. The acoustic machinery detection enables automated dynamic contextual information input without requiring manual user intervention.

[0072] Further, a method of sensing presence of a user in the space utilizes the auditory sensor in combination with presence sensors, thermal imaging sensors, camera 124, GPS location data, Wi- Fi / Bluetooth activity of user's devices, or combinations of these sensing modalities to collect and store timestamps in database 108 of when users were in the vicinity of the environment monitor 102. This multi-modal presence detection enables processing of data to calculate either individualuser or space occupant average exposure to pollutants. For example, health and safety standards recommend a time-weighted average of 8 hours for a working day to be below certain thresholds of pollution exposure. The data can be presented to users in a simplified method of TWA of the day, enabling users to take proactive action to ensure low pollution exposure.

[0073] In one embodiment, the auditory sensor is configured to measure sound levels in decibels (dB) to assess noise pollution in the monitored environment. In another embodiment, the auditory sensor comprises an ultrasonic sensor capable of detecting frequencies above 20 kilohertz, enabling detection of machine operation sounds while preserving user privacy by not capturing human speech frequencies. The auditory sensor may be implemented as a standalone component or integrated with other sensors in the environment monitor. The selection of the specific type of auditory sensor may be based on the contextual information of the space, the intended application, environmental conditions, or the specific acoustic phenomena to be monitored.

[0074] Further, a method of sensing presence of a user in the space, via a presence sensor, thermal imaging sensor, camera 124, or by deciphering user GPS location, Wi-Fi / Bluetooth activity of user’s phone nearby or a combination of these sensing data to collect store timestamps in a database 108 of when users were in the vicinity of the environment monitor 102. This enables processing of data, to calculate either a user’s or space occupants’ average exposure to pollutants in the space. For example, health and safety recommend a time-weighted average of 8 hours for a working day to be below certain thresholds of pollution exposure. The data can be presented to users in a simplified method of TWA of the day, enables users to take proactive action to ensure low pollution exposure.

[0075] Referring to FIG. 10 and FIG. 11, the user is recommended to add a new sensor to the environment monitor 102, based on the contextual information. For example, the contextual information analyzed that there is a risk of formaldehyde exposure in the space, but the environment monitor 102 does not monitor this pollutant. By advising the user to add a formaldehyde sensor module the monitoring can ensure tailored monitoring and advice to ensure full context awareness of the environment. The expansion 178 could be attached at the bottom of the main environment monitor 102 through connecting to expansion port 176. Alternatively, it can be a standalone formaldehyde monitoring module version of the environment monitoring system. Further, the expansion port 176 comprises a pogo pins. The pogo pins enables power and data transfer foradditional modules or alternative modes of powering. In some examples the expansion port comprises a power line, a ground line, and serial data (SDA), serial clock (SCL) lines for Inter- Integrated Circuit (i2c) communication. In another embodiment the expansion comprises a serial peripheral interface (SPI) communication. In another embodiment the expansion connects through the USBC port. In another embodiment sensor capability can be expanded wirelessly through networks and / or standards such as Bluetooth, Wi-Fi, Zigbee, LoRa, Matter.

[0076] In some examples, networked multiple environment monitor 102 give insights to a makerspace with multiple rooms. A user interface, for example on a tablet or desktop computing device, enables users to see multiple sensors of the space in one screen. The device can be colored to visually indicate spaces that may need further investigation to improve the environmental air quality.

[0077] Further, the environment monitor 102 comprises a Universal Serial Bus Type-C (USBC) port 156, a battery 168, an integrated magnet 170, and an inertial measurement unit (IMU). The USBC port 156 is an interface for charging and programming. In some examples, a battery 168 and inertial measurement unit (IMU) are integrated into the environment monitor 102 for portability. The battery 168 enables users to bring the environment monitor 102 closer to the activity and closer to the breathing zone of a user, to collect data that closely represents the air that the user is breathing. In some examples the environment monitor 102 has a switch 154 to turn off, to conserve battery 168. The environment monitor 102 could have a switch 154 to conserve battery life or change modes, including live data sampling, battery 168 conservation, Bluetooth low energy mode, Wi-Fi pairing mode, and standalone mode. Modes can include live data sampling at 3s intervals, conserved battery 168 mode sampling every 10 minutes (or specified time intervals), switching to a Bluetooth low energy mode, switching to a Wi-Fi pairing mode, switching to a standalone no network mode. The battery 168 could be rechargeable. Further, the battery 168 includes, but not limited to, lithium- ion battery. The alternative power supply could be solar power. Referring to FIG. 12, the magnet 170 enables secure attachment of the environment monitor 102 to metallic surfaces 180 in various environments.

[0078] In some examples, the IMU detects movement of the environment monitor 102, prompting the user to adjust the static or dynamic contextual information. Movement events are stored in adatabase 108 with a timestamp for later reference. The inertial measurement unit could be used to change the orientation of the e-ink display 144 from vertical viewing to horizontal viewing. Further, the inertial measurement unit enables quick access to functions in the environment monitor 102. For example, by tilting the environment monitor 102 by 90 degrees, the environment monitor 102 starts recording data at a higher sampling rate, and to label the data with a pre-set activity. This can be useful for context where the user frequently enters the activity. For example, on an electronic workbench, when a user starts to solder their electronics, they can use this function to record data during their activity at higher sampling rate, and automatically label the data without opening the software interface. In some examples, the environment monitor 102 could be mounted onto machines using the integrated magnet 170, allowing automated labelling of data with timestamps and machine usage status. The IMU data can detect normal operation of the machine and identify anomalies, such as in an air filtration device, sending alerts if pollution levels are not dropping as expected. The IMU data can be passed through a machine learning algorithm to classify whether a machine is in normal operation. The data can also be used to detect anomaly of machine usage. In some examples, this data is stored as contextual information for the environment monitoring system to enhance granularity in the recommendations. This can be advantageous for example if mounted onto an air filtration device, alerts can be sent to users if pollution levels are not dropping and inertial measurement unit data from the environment monitor 102 mounted on the air filtration device indicates abnormal operation state.

[0079] In some examples, the environment monitor 102 comprises of an additional microphone. Audio signals processed through machine learning algorithms to train and classify specific machines operating in space. In some examples, the microphone is capable of picking up ultrasound frequencies, detecting above 20 kilohertz. This enables additional contextual information through automated, labelling of data on the usage of hand tools like drills, Dremel® tools, and 3D printing machines, in the space whilst preserving privacy to users as it does not pick up human speech frequencies of sound. In some examples the sound level changes are recorded as an event with a timestamp, for ease of tagging the associated event by the user.

[0080] In some examples, the data collected from the sensors 174 of environment monitor 102 could be classified to detect if pollution levels are not normal. For example, a machine learning algorithm can be trained to classify what a typical laser cutter operation pollution data profilesfollow over a specified time window. If wrong materials are cut, pollution profiles will differ and an operator could be alerted. In some examples an auto-correlation algorithm with a specified time window is tuned to detect and forecast changes to pollution levels that is an anomaly for the context, and facilitate tagging of the data to provide context to the environment monitoring system.

[0081] In some examples, the two environment monitors 102 could be used to collect data to infer environmental conditions of a large space. One environment monitor 102 will be in a fixed position over the duration of the data collection for this inference algorithm. The second sensor could be placed in a second room in the same building over a duration of time. In one example, the duration could be one week, enabling collection of data with varying outdoor weather and environmental conditions. The second sensor can then be moved to another space over a similar duration to collect data. The data can then be trained using a machine learning algorithm, including Random Forest or LSTM neural networks, to estimate the environmental condition in the rooms that the second sensor collected data from. The Random Forest algorithm creates multiple decision trees using bootstrapped samples to predict environmental conditions, utilizing feature vectors comprising temporal measurements, contextual activity data, and external environmental factors. The LSTM neural network captures temporal dependencies in sequential environmental data through memory cells that learn daily cycles, weekly patterns, and activity-based fluctuations. The system constructs spatial correlation models that quantify how environmental conditions propagate between rooms, accounting for factors including air exchange rates, time delays, ventilation system operations, and physical barriers such as doors and walls. The machine learning models learn transfer functions that correlate measurements from the fixed monitor location with corresponding conditions in unmeasured spaces, enabling prediction of temperature, humidity, particulate matter, carbon dioxide, and volatile organic compounds in rooms without permanent sensor installation. The environmental condition, includes but not limited to, temperature, humidity, particulate matter, and carbon dioxide. The trained models enable real-time estimation of environmental conditions throughout the building using only the fixed monitor data, reducing sensor deployment costs while maintaining comprehensive environmental monitoring coverage. The system provides prediction confidence metrics and automatically triggers mobile sensor verification when prediction uncertainty exceeds acceptable thresholds.

[0082] The data collected of a spatial distribution of pollutants through a portable air quality sensor could be used to augment air quality data within a space. For example, using a head mounted augmented reality device with camera tracking capabilities, a user holding an air quality monitor in the visible scope can track air quality data whilst simultaneously augmenting the air quality data in space, using spatial positioning algorithms of the augmented reality device. Recommendations to improve the space to reduce health are presented to the user spatially, through the augmented reality interface. In some examples the augmented reality device has an integrated Al system and hardware enabling context information feature extraction from the images from the on-device camera. This allows for real time recommendations, for example if it sees a hardwood dust in the image, the user can be alerted whether the dust produced from machining the wood can be harmful to the user or not.

[0083] In some examples, the environment monitor 102 can be connected to a networked air filtration system or ventilation system. The environment monitor 102 data could be used to trigger solution systems to be powered on. This drives power optimization so that solutions can be automatically powered down if the environment is maintained at a good pollution level for a specified duration and / or there is no ongoing activity in the space. In another embodiment, the environment monitor 102 data can be processed with a machine learning algorithm to predict demand in the next hour to proactively drive automation systems before pollution levels are triggered.

[0084] In some examples, the environment monitor 102 is connected to a smart home network. The smart home network involves homeKit, home assistant, Matter, Alexa smart home, If This Then That (IFTTT) or ESPHome for a smart networked device. Alerts can be communicated through apple home pod via voice commands and voice feedback. In some examples, the environment monitor 102 displays a QR code to connect the environment monitor 102 with a mobile device or computing device. In some examples the environment monitor 102 could be connected to more than one user’s mobile device or computing device. Data can be easily shared with other users through a URL link to a website. Raw unprocessed data could be downloaded by users to analyze data independently.

[0085] FIG. 14 exemplarily illustrates a flowchart of a method 1400 for monitoring air quality and providing recommendation for managing air quality, according to some examples of the present invention. At step 1402, the user inputs contextual information. The contextual information comprises details related to, the activity that occurs in the space, tools available in the space, materials handled in the space, ventilation systems in the space, presence of air filtration devices, types of air filtration devices, whether the space is a shared space, the geographic location of the space and options to add further contextual information. As described earlier, the method of input of contextual information could be in different formats depending on the embodiment, varying from natural language interfaces, image content feature analysis with a machine learning algorithm processing a photo captured by the user, selection of a list of options from a list of activity or room types on a computing device 112. The contextual information could be stored on the device 112 or on a remote database 108.

[0086] At step 1404, the environment monitor 102 provides a user with recommendation based on the context information via a machine interface including a computing device 112. The recommendation is communicated to a user. The recommendation helps to raise awareness of a user around the risks of pollutants in the environment. The recommendation could be built through a decision tree matrix and / or a natural language, machine learning algorithm, trained on a dataset of pollution and health risks. One example of this dataset is the EH40 / 2005 Workplace exposure limits document published by the Health and Safety England, where each pollutant has details about the guidelines on legal limits to time- weighted-average pollution exposure, a 15 minute short-exposure limit, a CAS number indicating entries that are not part of approved workplace exposure limits (WEL), whether the pollutant is a carcinogen or not, whether the pollutant can be absorbed through the skin, whether the pollutant is capable of causing occupational asthma. One example of a recommendation is for users who entered a contextual information that the space is used for ceramics, the user could be recommended to take precautionary measures to reduce particulate matter levels, as clay dust might contain silica, which is a carcinogen. A beginner or casual hobbyist getting into ceramics may not know about these health risks. Alert thresholds for users to be notified could be adjusted according to the space contextual information. The granularity of information provided to the user can be adjusted according to the level of knowledge a user has about pollution and their workspace, which can improve over time. This could be evaluated through prolongedengagement with the user interface, and / or a short quiz to evaluate knowledge around air pollution within their context.

[0087] At step 1406, the environment monitor 102 suggest changes to alert profiles and recommendations. The user could be prompted to change their user interface for their specific contextual information via machine interface. For example, if a user is in a ceramics space, they would have the option to see their time-weighted-average data next to the reference guidance extracted from a relevant local authority advising health and safety. The interface can pre-filter the dynamic contextual information to provide selections of dynamic activities or most likely candidates of activities occurring in the space. For example, firing a could be a first option guiding users to enter dynamic contextual information. At step 1408, the system processes the environmental monitor data and contextual information to tailor recommendations. Data collected from the environment monitor 102 is processed to provide analytics 126 and recommendations to the user on ways to improve and maintain good air. Analytics 126 can be a processing of the collected data of a time weighted average of the data during working hours, set by the user. A monthly average of the time weighted average can also be shown. If, in step 1402, a geographic location was entered, local weather and pollution forecasts could be integrated into the recommendation and live display of the user interface.

[0088] In some examples, the e-ink screen 144 on the environment monitor 102 could be used to display custom information. The custom information includes, but not limited to, a clock, a local weather forecast, an outdoor temperature, an action recommendation, an air quality data from the integrated sensors, or a calendar schedule. The user could integrate their calendar, and display the current schedule and next upcoming schedule in the calendar. The interface could be connected with a local public transport timetable API, to display the time of next departure of a bus nearby.

[0089] If the environmental data over a long duration does not correlate with the statistical distribution profile of a typical operational data in the environment, the users can be alerted. This might be due to obstructions at the inlet of the sensor inlet airflow, orientation of the environment monitor 102, or deterioration of the sensor due to physical damage, electronic hardware damage or end of life of the sensor. The user can be advised to get maintenance for the environment monitor 102.

[0090] For example, the user inputs contextual information related to the space, for example, ceramic studio. The contextual information comprises the details regarding activity, machine, materials, ventilation, air filtration, heating and cooling, and other details. The activity includes, but not limited to, a pottery, a glazing, a kiln firing, a turntable. The machine includes, but not limited to, a kiln and a turntable. The materials, includes, but not limited to, a clay. The ventilation system could be windows. The air filtration could be ducted fan. Heating and cooling are none. The other details could be regarding shared space. The contextual information is stored with a timestamp and the data is analyzed. Based on the analysis, the system determines that the clay work in pottery might have harmful particles of silica, that can lead to silicosis. Further, the environment monitor 102 measures PM2.5 and records the activity and materials used in the space, and health risk estimates. At step 1508, the system provides recommendations to user based on the context information. For example, the recommendations include working 8-hour average, suggest to change alert settings to notify if any parameters exceed advice. If 8 hour working average exceeded advice, the system provides suggestions to integrate an air filtration with HEPA grade filtration. Further, the links to suggested products are provided as URL.

[0091] FIG. 15 exemplarily illustrates the flowchart of a method 1500 for providing recommendation for managing air quality, according to some examples of the present invention. At step 1502, the user inputs a dynamic contextual information. The contextual information could be starting of a kiln firing and the duration of 8 hours. At step 1504, the dynamic contextual information is stored in a database 108 along with a time stamp, user ID and device ID. At step 1506, the data could be analyzed, using a machine learning algorithm connected to a natural language interface to communicate with users on the details of the pollution profile during the duration of the contextual information event. At step 1508, the analyzed and processed contextual information converted into natural language is communicated to user, with a format tailored to the user profile. Recommendations can be adjusted. The system can learn and rank the users air pollution comprehension level as users interact with the system.

[0092] In some examples, a large language model (LLM) trained on datasets including health advice, creative workspace safety best practices, facility manager training, and health and safety standards provides analytical insights. The LLM interfaces with the database 108 via a naturallanguage interface to deliver real-time, personalized recommendations to users. This system allows for dynamic analysis and guidance based on contextual data, enhancing safety and usability in various environments.

[0093] The LLM's unique capability stems from its integration with real-time environmental sensor data and contextual workspace information, creating capabilities that standalone Al systems cannot match. The system processes live air quality measurements alongside user-provided context to generate situationally relevant health and safety advice that goes beyond simple threshold detection.

[0094] For example, when PM2.5 levels spike during hardwood sanding activities, the system provides material-specific guidance on the software recognizing that certain hardwood species are carcinogenic, requiring enhanced protective measures. The LLM correlates real-time pollution data with material hazard profiles to provide targeted advice on ventilation enhancement, filtration requirements, respiratory protection, and specialized cleanup procedures. The system further enriches recommendations with statistical insights regarding health outcomes, allergic reactions, and respiratory issues associated with specific materials based on epidemiological data.

[0095] The system generates contextually appropriate recommendations through a multi-stage reasoning process focused on long-term health protection and cumulative exposure management. First, contextual pollution analysis is performed, wherein real-time sensor readings are correlated with specific activities and materials to identify hazard-specific exposure patterns rather than relying on generic particle detection. Second, a material- specific health risk assessment is applied, using occupational health guidelines enhanced with material-specific carcinogenicity data, respiratory sensitization potential, and long-term health implications. Third, predictive pattern recognition is carried out by utilizing Al-driven analysis to identify pollution patterns that correlate with specific activities, enabling detection of equipment failures, maintenance needs, and abnormal exposure scenarios. Fourth, behavioral modification guidance is generated, providing motivation- focused recommendations that explain the health importance of protective measures to encourage behavior adoption.

[0096] The system also employs sophisticated Al analysis to identify pollution patterns and equipment performance indicators that would otherwise be undetectable. Equipment failuredetection is achieved by analyzing air pollution data patterns to identify malfunctions, such as failures in 3D printers, thereby providing early warning systems and failure prevention guidance. For maintenance optimization, the system predicts when dust extraction filters are approaching capacity, thus preventing occupant exposure due to reduced filtration effectiveness. Ventilation effectiveness assessment is performed by measuring real-time performance through pollution clearance rates and air exchange effectiveness. Anomaly detection algorithms are implemented to identify unusual pollutant patterns that deviate from established weekly profiles, alerting users to investigate potential new sources of pollution or equipment issues. Additionally, automated integration guidance is provided, offering step-by-step instructions for integrating air purification systems with monitoring data to support automated response protocols.

[0097] The system processes contextual information from diverse input formats using a flexible data schema designed to accommodate varied information types. It performs flexible data integration across broad categories, including activity type, materials in use, equipment operation status, environmental conditions, and user profile characteristics, without requiring strict data structure compliance. The system conducts document and media analysis, processing text documents, photographs, videos, and audio recordings with automatic categorization based on context relevance and safety implications. Furthermore, it integrates cumulative exposure data by interfacing with air quality monitoring systems to receive processed cumulative exposure information, time-weighted averages, and dose accumulation models for LLM analysis, without performing exposure calculations directly.

[0098] The LLM implementation supports flexible deployment scenarios. It offers cloud-based processing for full-featured analysis using comprehensive model capabilities to conduct complex, multi-factor health risk assessments. Edge computing optimization is also supported through reduced-parameter models that are suitable for edge devices, optimized for real-time audio feedback and basic recommendation generation. Offline functionality is enabled via local storage of materialspecific safety protocols and activity-based guidelines, allowing the system to provide enhanced recommendations beyond basic ventilation and air purification advice. The enhanced offline guidance includes activity-specific risk explanations, motivational health education, and behavioral modification strategies to encourage protective practices even when there is no network connectivity.

[0099] The system develops unique workspace pollution fingerprints through extended data collection and analysis. After accumulating at least three days of environmental data, the system begins generating daily, weekly, monthly, annual, and sliding window anomaly comparisons, thereby establishing baseline pollution patterns specific to each workspace and sensor configuration. Seasonal and long-term pattern recognition is also performed by continuously collecting data throughout annual cycles to identify seasonal variations, usage pattern changes, and equipment degradation signatures. The system further supports external data integration by correlating workspace, home, and general indoor environment pollution patterns with public datasets, including local air quality station data, weather trends, satellite air quality data, and private environmental datasets. This enables the system to distinguish indoor pollution from outdoor sources. Contextual pattern evolution is achieved by combining pollution fingerprints with user- inputted contextual information to refine understanding of activity-specific exposure patterns and environmental responses over time.

[0100] The system incorporates longitudinal health tracking capabilities to enable predictive health modeling and symptom correlation. Cumulative exposure health prediction is carried out using algorithms that forecast long-term health outcomes based on cumulative exposure patterns, material-specific risk factors, and individual susceptibility profiles. Users are enabled to track health symptoms and other indicators as part of contextual data collection through health journaling integration. Exposure-symptom correlation analysis is performed by correlating reported health symptoms with historical exposure data to identify personal sensitivity patterns and refine individual risk assessments. Predictive health risk modeling is implemented using machine learning algorithms to forecast potential health impacts based on continued exposure patterns and to recommend preventive interventions.

[0101] The system also provides location-specific recommendations based on regional environmental factors. Regional risk factor integration is achieved by adapting recommendations based on geographic location, incorporating considerations such as crop burning seasons, industrial proximity, construction activity, and pesticide usage in agricultural regions. Geolocation-based contextual advice is generated by utilizing GPS and other location data to provide region-specific guidance for outdoor pollution infiltration, seasonal air quality variations, and compliance with localregulatory requirements. Cultural and regulatory adaptation is performed by adjusting recommendation language, safety priorities, and compliance instructions based on local regulatory frameworks and cultural safety practices across multiple regions.

[0102] The system incorporates systematic approaches for managing ambiguous or conflicting information. Sensor-context conflict resolution is carried out when sensor data contradicts user- provided contextual information; in such cases, the system prompts users for additional input or verification to resolve discrepancies. Uncertainty quantification is implemented by assigning confidence levels to recommendations based on data quality, context completeness, and historical validation. Low-confidence scenarios trigger user feedback requests. Adaptive response protocols are developed for handling ambiguous safety situations, including graduated response strategies and conservative default recommendations in cases of uncertainty.

[0103] The system generates unique intellectual property through proprietary dataset development and competitive differentiation. A recommendation performance database is maintained, aggregating user feedback on recommendation effectiveness to support model optimization. A curated expert knowledge base is developed using consultations, field studies, and validation by professionals. A real-world exposure pattern library is built by collecting workspacespecific pollution profiles, equipment failure signatures, and data on intervention effectiveness. Contextual intelligence algorithms are developed for correlating environmental sensor data with health outcomes, equipment performance, and behavioral modifications.

[0104] To ensure current and accurate safety guidance, the system supports frequent knowledge base updates, incorporating new regulatory changes, research findings, and safety bulletins, all subject to validation and testing. Real-time integration of authoritative safety alerts and regulatory updates is also supported. Version control mechanisms are implemented to maintain compliance with current standards across various regions. Continuous integration of peer-reviewed research and industry best practices is employed to enhance recommendation quality.

[0105] The system includes adaptive learning and user feedback integration mechanisms for continuous improvement. Recommendation format optimization allows users to select preferred communication styles, with this feedback informing future personalization. Performance ratingsystems enable users to rate the relevance and actionability of recommendations, which is used in model retraining. Crowd-sourced validation aggregates feedback across installations to validate recommendation effectiveness and identify areas for refinement. Expert escalation protocols automatically flag scenarios for expert review, including those involving novel material combinations, conflicting health guidelines, user-reported symptoms, or low-confidence recommendations .

[0106] Validation is provided by certified professionals, including industrial hygienists, occupational health experts, medical doctors, environmental health researchers, and regulatory representatives. Quality assurance mechanisms include confidence scoring for all generated recommendations, automatic flagging of uncertain or novel scenarios, feedback rating systems, continuous improvement protocols, and network learning effects from multiple deployments, while maintaining user privacy protections.

[0107] The system also features predictive risk modeling by leveraging historical and environmental data to forecast potential exposure scenarios based on planned activities and environmental conditions. It generates proactive recommendations before hazardous conditions arise, models cumulative exposure effects over time, and predicts equipment maintenance needs based on usage patterns and environmental stress factors.

[0108] To support inclusivity and accessibility, the system provides multi-language support for safety communications in diverse work environments. Audio output capabilities are included for visually impaired users, and simplified language modes are provided for users with varying literacy levels. Visual diagram generation supplements textual recommendations to enhance comprehension.

[0109] Finally, the system ensures regulatory compliance by maintaining current knowledge of applicable local, national, and international safety regulations. It aligns with industry-specific safety standards for various creative and industrial workspaces, integrates insurance -related safety requirements, and supports documentation for compliance reporting and audit trails.

[0110] The LLM system is designed to achieve a response time of less than 3 seconds for standard recommendation requests. It delivers an accuracy rate of 98% or higher for material identification and risk assessment when adequate contextual information is provided. The system targets user satisfaction scores of 4.5 or higher (out of 5) for recommendation relevance and clarity. It ensures compliance with relevant data privacy and security standards, including GDPR and HIPAA where applicable. The system also achieves a successful anomaly detection rate of 95% or higher for equipment failure and unusual exposure scenarios. This expanded specification provides a comprehensive framework for implementing advanced Al capabilities while maintaining focus on practical, safety-critical applications in environmental monitoring and workplace health protection.

[0111] FIG. 16 exemplarily illustrates a screenshot 1600 of a user interface to enable a user to select information regarding the environment of a workspace or creative space, according to some examples of the present invention. The screenshot 1600 is regarding "ambient one'' (ambient one is the name of an air quality monitor, produced by “Ambient Works”). Further, the screenshot 1600 displays the details of the ceramics space. The details include the information regarding air quality, activity involved in the space, particulate matter (PM), volatile organic compounds (VOC), carbon dioxide (CO2) and tips and advices. The screen shot displays that the air quality is poor. The PM 2.5 is 78 pg / m3and PM 10 is 72 pg / m3. The VOC is 98 index and CO2 is 456 ppm. Further, the user interface displays the activities involved in ceramics workshop such as firing, glazing, and clay handling, which the user may select. Additionally, the screenshot 1600 displays tips and advice related to pollution risks in the contextual information stating that silicosis is an irreversible lung disease cause by inhaling silica, that is often produced from clay work activity in ceramics. Further, the screen shot has an “improve air quality” option.

[0112] FIG. 17 exemplarily illustrates a screenshot 1700 of a user interface displaying a real time data and a button for dynamic data labelling, according to some examples of the present invention. The screenshot 1700 is regarding batch space. Further, the screenshot 1700 displays the data of the ceramics space. The data include the information regarding air quality, environment of ceramics workshop, and a data of a day, a week, and a month. The screen shot displays data of today that the air quality is poor and PM 2.5 is 78 pg / m3. Additionally, the screenshot 1700 displays the environment of ceramics workshop, that silicosis is an irreversible lung disease caused by inhalingsilica, that is often produced from clay work activity in ceramics. Further, the screen shot has a “record activity” option.

[0113] FIG. 18 exemplarily illustrates the screenshot 1800 of a user interface displaying recommendations, according to some examples of the present invention. The screenshot 1800 is regarding notification center. The notification center includes the alert sent by the system. The alert on notification center displays the PM 2.5 Level Rising-Sensor. Further, the screenshot 1800 displays the rising level of the pollutant. The alert sent at 03:22 PM shows “Get fresh air with an air purifier, open the window or change, what you’re doing. Pollutant level is 50”. Further, the alert sent at 03:27 PM shows “Get fresh air with an air purifier, open the window or change, what you’re doing. Pollutant level is 10.5”. Further, the screenshot 1800 has a notification showing monocle on design, “Extra: Kara Pecknold” is Available.

[0114] FIG. 19 exemplarily illustrates the screenshot 1900 of a user interface displaying the details of the workspace or creative space, according to some examples of the present invention. The screenshot 1900 displays a floor plan of the work space. The floor plan includes, a total dimension, an area identified, and an extraction. The screenshot 1900 displays the total dimension is 168 sq.ft., the identified areas are 4, and extraction is not present. Further, the screenshot 1900 has options that includes “scan new space”, “view 3D model” and “edit floor plan”.

[0115] FIG. 20 exemplarily illustrates the screenshot 2000 of a user interface displaying analytics of an environment, according to some examples of the present invention. The screenshot 2000 displays the data related to the pollutants on daily average and 8hr time weighted average of a day. The daily average is 48 pg / m3and the 8hr time weighted average is 40 pg / m3. Further, the screenshot 2000 has a record and an advice option. The screenshot 2000 further has a data of a day, a week, and a month.

[0116] FIG. 21 exemplarily illustrates the screenshot 2100 of a user interface displaying a real time data and prompts for action, according to some examples of the present invention. The screenshot 2100 is regarding the wood workshop. Further, the screenshot 2100 displays a data of a day, a week and a month. Further, the screenshot 2100 displays the air quality is poor and high PM2.5, which is 78 pg / m3. Further, the screenshot 2100 display a question, like, “Are you wearing a mask?”, “Is your extractor on?”. Further, the screen has option to “Record activity”.

[0117] FIG. 22 exemplarily illustrates the screenshot 2200 of a user interface displaying a context aware recommendation to the users, according to an embodiment of the present invention. The screenshot 2200 displays advice for the ceramic workshop environment. The advice is regarding a material, a process and an activity. The material could be clay. Further, the system displays advice that silicosis is an irreversible lung disease cause by inhaling silica, that is often produced from clay work activity in ceramics. The processes include dipping- glazing. Further, the system advices to check health risks of glazing substance and ensure you have the right protective equipment with an option “More info on glazing”. The activities include mopping. Further, the advice includes cleaning the studio with precaution.

[0118] FIG. 23 exemplarily illustrates the screenshot 2300 of a user interface displaying an activity labelling of data involved in firing interface and the high PM2.5 levels indicating poor air quality, according to some examples of the present invention. The graph provides an overview of particulate matter (PM) levels in the ceramic studio during the firing activity between 8 am to 8pm. The PM is 53.5 pg / m3, which might contain silica, alumina, and metal oxides. Further, the daily average of PM detected is 27.5.

[0119] FIG. 24 exemplarily illustrates the screenshot 2400 of a user interface displaying an activity labelling of data involved in throwing and low PM2.5 levels indicating good air quality, according to some examples of the present invention. The graph provides an overview of particulate matter (PM) levels in the ceramic studio during the throwing activity. The screenshot 2400 displays a good air quality is 9.0 pg / m3, which might contain silica, alumina, and metal oxides. Further, the daily average of PM detected is 27.5.

[0120] FIG. 25 exemplarily illustrates a screenshot 2500 of a user interface providing data visualization of particulate matter (PM) concentration related to specific activities in a ceramic studio, according to some examples of the present invention. The graph provides an overview of particulate matter (PM) levels in the ceramic studio during the glazing activity, and highlights theinfluence of sanding and painting on air quality at different times of the day. The PM levels is 42 mg / m3.

[0121] According to the present invention, candidate item recommendations refer to recommendations that are preliminarily identified as potentially relevant based on an analysis of contextual information and the user’s profile. These candidate recommendations represent a filtered subset of all available recommendations, selected specifically to align with the identified space type and user categories.

[0122] The system utilizes a recommendation template database that is populated from a variety of authoritative sources. These sources include, but are not limited to, governmental regulatory bodies such as the Environmental Protection Agency (EPA), Occupational Safety and Health Administration (OSHA), National Institute for Occupational Safety and Health (NIOSH), World Health Organization (WHO), and International Labour Organization (ILO); medical and public health organizations including respiratory health associations, occupational medicine societies, and cancer research institutions; peer-reviewed academic research publications and systematic literature reviews; professional standards-setting bodies such as the American Society of Heating, Refrigerating and Air-Conditioning Engineers (ASHRAE), ASTM International, and the International Organization for Standardization (ISO); as well as domain experts in specific safety fields including art materials safety, woodworking safety, ceramics safety, and chemical safety. Additionally, the database incorporates established references such as toxicology databases, chemical safety data sheets (MSDS), and environmental health guidelines. The recommendation templates are periodically updated by the system to reflect newly published research findings, updated regulatory guidance, and evolving best practices from these authoritative sources.

[0123] The processor is configured to generate candidate item recommendations using at least one of the following techniques: (a) rule-based selection mechanisms employing decision tree logic to match contextual inputs with predefined templates; (b) machine learning classification models that categorize the context and select from predefined recommendation sets; or (c) generative language models that formulate context-specific recommendations by processing the contextual data through neural networks trained on environmental health and safety corpora.INTELLIGENT RECOMMENDATION GENERATION SYSTEM

[0124] The processor generates candidate recommendations through systematic analysis of contextual information using at least one of: decision tree filtering, machine learning classification, or generative artificial intelligence models. When using generative models, the system employs natural language generation architectures including transformer-based models, attention mechanisms, or neural networks trained on environmental health and safety datasets to create contextually appropriate recommendations. The generative models analyze the environmental monitoring data, space characteristics, ongoing activities, and user risk profiles to produce human- readable recommendations that may include equipment suggestions, safety protocols, exposure mitigation strategies, and health risk warnings tailored to the specific environmental conditions and user context. The generative models are trained or fine-tuned using datasets comprising regulatory guidelines, occupational health standards, research-based health impact studies, and domainspecific safety protocols to ensure accuracy and relevance of generated recommendations.

[0125] The system generates candidate item recommendations by analyzing the contextual information against a database of recommendation templates. Each template is associated with specific environmental conditions, space types, user risk profiles, ongoing activities in the space, and objects present in the space that are known sources of pollutants. The processor evaluates multiple potential recommendations and designates them as "candidates" when they meet initial relevance criteria based on the contextual information. The candidate recommendations comprise a filtered subset of applicable recommendations tailored to the specific space type, ongoing activities, detected environmental conditions, and user categories, wherein the filtering process ensures that only contextually relevant recommendations are considered for presentation to the user.

[0126] The database comprises risk profiles of contextual information sourced from authoritative sources including government regulatory agencies such as EPA, OSHA, NIOSH, WHO, and ILO; medical and health organizations including respiratory health societies, occupational medicine associations, and cancer research institutes; peer-reviewed academic research publications and systematic reviews; professional standards organizations such as ASHRAE, ASTM, and ISO; specialized domain experts in areas including art materials safety, woodworking safety, ceramics safety, and chemical safety; and established databases including toxicology databases, chemical safety data sheets, and environmental health guidelines. The risk profiles include documented healtheffects of particulate matter on respiratory systems, cardiovascular health, neurological function, and cancer risk. For example, the database contains research-based risk data indicating that PM2.5 exposure affects lung function, increases risk of asthma exacerbation, contributes to cardiovascular disease, may impact brain health and cognitive function, and is associated with increased cancer risk. The risk profiles further include volatile organic compound exposure data indicating respiratory irritation, neurological effects, and carcinogenic potential, formaldehyde exposure data indicating respiratory irritation, asthma triggers, and cancer risk, and carbon monoxide exposure data indicating oxygen deprivation effects and neurological impacts.

[0127] The recommendation templates include activity- specific templates corresponding to ongoing activities such as kiln firing, wood turning, metal casting, soldering, varnishing, gluing, photo development, 3D printing, laser cutting, welding, grinding, sanding, painting, screen printing, jewelry making, glassblowing, and chemical mixing. The templates further include object-specific templates corresponding to pollutant-emitting objects present in the space, including but not limited to furniture that may off-gas formaldehyde, adhesives and glues that emit volatile organic compounds, candles that release particulate matter and potentially harmful compounds, resins that emit volatile organic compounds during curing, cleaning supplies that release chemical vapors, paints and solvents that emit volatile organic compounds, electronic equipment that may emit ozone or other compounds, printers and copiers that emit particulate matter and ozone, carpets and textiles that may harbor allergens and emit VOCs, and building materials that may off-gas various chemicals.

[0128] Each recommendation template incorporates the research-based risk profile data to specify threshold criteria for environmental sensor readings, duration of exposure limits based on documented health impacts, and user-specific risk factors including age, pre-existing respiratory conditions, cardiovascular health status, pregnancy status, immune system compromised conditions, and occupational exposure history. The processor matches the current contextual information including space type, ongoing activities, identified pollutant-emitting objects, and applicable health risk factors against the template criteria to generate relevant candidate recommendations prioritized by severity of documented health impacts. The prioritization algorithm weighs factors including immediacy of health risk, severity of potential health outcomes, user vulnerability factors, and availability of mitigation measures.

[0129] The decision tree filtering process evaluates contextual parameters in hierarchical order:(1) space type classification determining the primary environmental hazards associated with the space, (2) activity risk assessment analyzing ongoing or planned activities that may generate pollutants, (3) pollutant source identification cataloging objects and materials present that are known emission sources, (4) user vulnerability factors assessing individual risk characteristics, and (5) current environmental thresholds comparing real-time sensor data against established exposure limits. Each decision node applies weighted scoring based on documented health impact severity from the research-based risk profiles, creating a prioritized list of candidate recommendations. The decision tree structure enables rapid filtering of the recommendation database to identify the most relevant subset of recommendations for the current context.

[0130] When employing generative artificial intelligence models, the system utilizes transformer-based architectures trained on curated datasets of occupational health guidelines, peer- reviewed research abstracts on environmental health impacts, regulatory exposure limits from OSHA, EPA, WHO, NIOSH, and international health organizations, and safety protocols from industry-specific organizations covering woodworking, metalworking, ceramics, electronics, chemical handling, textile work, and other creative and industrial activities. The generative models are fine-tuned using domain-specific environmental health datasets to ensure recommendations are technically accurate, contextually appropriate, and actionable for users with varying levels of technical expertise.

[0131] For example, in a ceramic studio where the contextual information indicates clay handling activities and the environmental monitor detects elevated PM2.5 levels above 35 pg / m3, the system generates candidate recommendations including: (1) immediate respiratory protection recommendations prioritized as high severity due to silicosis risk from crystalline silica exposure,(2) ventilation improvement suggestions including exhaust fan installation or window opening protocols, (3) wet cleaning method recommendations to reduce airborne particle resuspension, (4) work practice modifications such as using pre-wetted clay or working in smaller batches, and (5) long-term monitoring recommendations for cumulative exposure tracking. The generative Al model produces natural language explanations such as "Clay dust may contain crystalline silica, which cancause permanent lung damage called silicosis. Consider wearing an N95 respirator rated for fine particles and using wet cleaning methods to reduce airborne particles."

[0132] In another example, for a woodworking space where the contextual information indicates hardwood machining activities and elevated PM 10 levels are detected, the system generates candidate recommendations prioritized by the carcinogenic potential of certain hardwood dusts. The recommendations may include species-specific guidance, such as "Beech, oak, and other hardwood dusts are classified as Group 1 carcinogens. Use local exhaust ventilation at the point of dust generation and wear appropriate respiratory protection rated for wood dust. " The system correlates the specific wood species input through contextual information with established carcinogenic classifications from IARC (International Agency for Research on Cancer) databases to provide targeted health risk warnings.

[0133] These candidates are then subject to the matching value and threshold analysis to determine which should be actively triggered and presented to the user based on real-time environmental monitoring data and calculated health risk levels derived from the research-based risk profiles. The matching value represents a quantitative assessment of how closely the current environmental conditions and contextual factors align with the applicability criteria for each candidate recommendation. The threshold analysis compares this matching value against predetermined recommendation strength thresholds, ensuring that only recommendations with sufficient relevance and potential impact are presented to avoid information overload while maintaining comprehensive safety coverage.

[0134] This multi-modal approach to recommendation generation enables the system to provide both immediate, rule -based safety alerts for well-established hazard scenarios and sophisticated, contextually-aware guidance that adapts to novel environmental conditions and user needs. The combination of template-based filtering and generative Al capabilities ensures comprehensive coverage of known hazards while maintaining flexibility to address emerging environmental health challenges and unique workspace configurations that may not be fully captured in predefined templates.

[0135] The system categorizes candidate recommendations into distinct presentation formats to ensure information is delivered in digestible, actionable units tailored to user comprehension and immediate needs. The categorization system organizes recommendations into at least one of: solution-based recommendations, risk-based recommendations, equipment recommendations, behavioral modification recommendations, and monitoring recommendations. Each category is presented as discrete information cards or interface elements designed to avoid cognitive overload while maintaining comprehensive safety coverage.

[0136] Solution-based recommendation cards present immediate actionable steps to address detected environmental conditions. For example, when elevated particulate matter is detected in a woodworking space, a solution-based card may display "Install Dust Extraction System" with specific guidance such as "Connect a shop vacuum with HEPA filtration directly to your table saw to capture 95% of generated dust at the source." These cards focus on implementable solutions rather than theoretical risks, providing users with clear next steps including equipment specifications, installation guidance, and expected effectiveness metrics derived from the researchbased risk profiles.

[0137] Risk-based recommendation cards present health hazard information coupled with specific mitigation requirements. For example, when formaldehyde is detected or contextual information indicates presence of formaldehyde-emitting materials, a risk-based card may display "Formaldehyde: Carcinogen Detected" with explanatory text such as "Formaldehyde is classified as a Group 1 carcinogen causing nasal and throat cancer. Use activated carbon filtration rated for formaldehyde removal and ensure ventilation rates of at least 6 air changes per hour." These cards combine hazard identification with targeted protection requirements, enabling users to understand both the severity of the risk and the specific protective measures needed.

[0138] Equipment recommendation cards provide specific product guidance with technical specifications matched to the detected environmental conditions and space characteristics. For example, when volatile organic compounds are detected above threshold levels in a painting studio, an equipment card may display "Air Purifier: VOC Removal Required" with specifications such as "Select air purifier with activated carbon filter rated for VOCs, minimum CADR of 200 CFM for your 400 sq ft space, and automatic air quality monitoring." The equipment recommendationsinclude sizing calculations based on space volume from contextual information, specific filter media requirements based on detected pollutants, and performance metrics to ensure adequate protection levels.

[0139] The presentation system adapts the complexity and technical detail of each recommendation card based on user profile characteristics including technical expertise level, previous system interactions, and stated preferences for information depth. Novice users receive simplified cards with basic explanations and clear action items, while expert users may receive detailed technical specifications, regulatory reference citations, and advanced configuration options. The system tracks user engagement with different card types and adjusts presentation formats to optimize comprehension and implementation rates, ensuring that critical safety information is effectively communicated regardless of user technical background.

[0140] The system implements a hierarchical threshold determination process to provide contextually appropriate environmental monitoring and recommendations. The recommendation strength threshold is determined through a multi-tiered approach that prioritizes regulatory compliance while allowing user customization for enhanced safety margins.

[0141] In one embodiment, the system automatically determines applicable threshold values based on the user's geographic location. Location identification may be accomplished through multiple methods including mobile device GPS coordinates, manual user entry within the application interface, user language selection, or IP geolocation techniques. Based on the determined location, the system accesses the corresponding national occupational health and safety recommendations, such as HSE workplace exposure limits in the United Kingdom, OSHA permissible exposure limits in the United States, or equivalent occupational health standards from other countries.

[0142] When national standards are unavailable or cannot be determined, the system implements a systematic fallback hierarchy to ensure comprehensive coverage. The system first references ILO (International Labour Organization) occupational exposure limits when available for specific pollutants, followed by WHO (World Health Organization) air quality guidelines for broader environmental standards. For pollutants where international data remains limited, the system mayreference ACGIH (American Conference of Governmental Industrial Hygienists) Threshold Limit Values, which are widely recognized and referenced internationally when local limits do not exist. This hierarchical fallback mechanism ensures that users receive appropriate health-protective guidance regardless of their location or the comprehensiveness of local regulatory standards.

[0143] The contextual information regarding workspace type significantly influences threshold selection and pollutant prioritization. For example, ceramics workshops may trigger monitoring for silica exposure limits due to clay dust, with the system referencing the applicable 50 pg / m3respirable crystalline silica standard as an 8-hour time-weighted average. Woodworking spaces may reference exotic hardwood dust standards, and office environments may focus on formaldehyde limits from carpets or printing equipment. The system categorizes pollutant-specific advice into digestible card or tab formats, allowing users to examine each pollutant risk separately while understanding the specific health impacts relevant to their workspace context.

[0144] The matching value calculation represents the ratio between measured environmental conditions and the applicable threshold limit. For example, if PM2.5 measurements show 30 pg / m3and the applicable WHO guideline threshold is 15 pg / m3, the matching value would be calculated as 2.0. The recommendation strength threshold may be configured at various trigger points, such as 0.8, meaning recommendations are generated when measured values reach 80% of the applicable limit, enabling proactive health protection before regulatory limits are exceeded.

[0145] Advanced users may customize threshold settings to establish additional safety margins beyond regulatory requirements. For pollutants with established legal workplace limits, such as crystalline silica or carbon monoxide, users may set supplementary alert thresholds below the regulatory standard but cannot override the legal limits themselves. For pollutants such as CO2 and PM2.5 levels in non-regulated environments, users may fully customize threshold values and trigger ratios according to their specific sensitivity requirements or operational preferences. Users may also manually select specific concentration levels to trigger the generation of contextual advice and recommendations .Matching Value Calculation and Processing:

[0146] The processor-readable instructions further determine a matching value for comparison against recommendation strength thresholds to trigger appropriate user recommendations. The matching value comprises measured pollutant concentrations obtained from one or more environment monitoring sensors, processed and analyzed according to the specific pollutant type, measurement context, and user exposure scenarios.Measurement Types and Units

[0147] The matching value may comprise various pollutant concentration measurements including, but not limited to: particulate matter concentrations expressed in micrograms per cubic meter (pg / m3) for PM2.5, PM10, and ultrafine particles; gaseous pollutant levels in parts per million (ppm) or parts per billion (ppb) for carbon monoxide (CO), carbon dioxide (CO2), nitrogen dioxide (NO2), and ozone (O3); volatile organic compound readings as total VOC index values or specific compound concentrations; formaldehyde levels in milligrams per cubic meter (mg / m3); radon concentrations in becquerels per cubic meter (Bq / m3); and sound levels in decibels (dB) for noise pollution assessment. The specific units and measurement ranges are automatically selected based on the sensor type and detected pollutant characteristics.Temporal Processing Methods

[0148] The matching value calculation may involve various temporal processing approaches depending on the applicable safety standards and exposure assessment requirements. Real-time instantaneous readings provide immediate hazard detection for pollutants with acute toxicity effects, such as carbon monoxide or hydrogen sulfide. Time-weighted averages calculated over 8-hour work periods align with occupational exposure standards, while 15-minute short-term exposure limits capture peak exposure events. Rolling averages computed over user-defined periods including 1- hour, 4-hour, and daily intervals enable trend analysis and long-term exposure monitoring. Peak detection values capture maximum exposure events for documentation and risk assessment purposes. Cumulative exposure calculations provide dose-based assessments particularly relevant for carcinogenic substances like silica dust or asbestos fibers.Contextual Adjustment Factors

[0149] The matching value may be contextually adjusted based on space-specific factors, user activities, and environmental conditions. Breathing zone proximity weighting applies increasedsensitivity when sensors are positioned near user activity areas, recognizing that pollutant concentrations may vary significantly within a workspace. Spatial interpolation algorithms estimate exposure levels between multiple sensors in larger workshop spaces, accounting for air circulation patterns and pollutant dispersion characteristics. Activity-specific multiplication factors automatically adjust sensitivity during high-risk operations, such as increased particulate monitoring during woodworking or enhanced VOC detection during painting activities. Materialspecific risk coefficients modify threshold values based on the carcinogenicity, asthmogenic potential, or acute toxicity profiles of materials being processed in the workspace. Ventilation effectiveness corrections account for air circulation patterns, filtration system performance, and seasonal variations in natural ventilation.

[0150] The system further incorporates particulate matter composition analysis to enhance exposure assessment accuracy for specific hazardous components. The processor-readable instructions may utilize default research-based proportion values of PM2.5 composition derived from established literature or databases, enabling initial risk assessment for common materials and activities. Alternatively, actual measured concentration composition data may be entered into the air quality monitoring system through user input interfaces or integrated with external analytical equipment results. This compositional analysis facilitates customized trigger thresholds for specific activities with specific materials, allowing the system to apply component- specific exposure limits rather than generic particulate matter standards. For example, in ceramics workspaces, the system may estimate silica exposure levels within measured particulate matter to apply silica-specific occupational exposure limits. The processor-readable instructions may access material-specific composition databases and enable users to input custom composition data for specialized materials or novel applications, thereby providing enhanced health risk assessments for complex multicomponent exposures common in creative and industrial workspaces.Risk Assessment Integration

[0151] The matching value may be processed through comprehensive risk assessment algorithms that incorporate user-specific factors and exposure scenarios. Health risk multipliers adjust threshold sensitivity based on user demographic profiles, including age, pre-existing respiratory conditions, pregnancy status, or other vulnerability factors. Exposure duration weighting differentiates between full-time workers and occasional users, applying appropriate safety factorsfor different exposure scenarios. Synergistic effect calculations account for the presence of multiple pollutants simultaneously, recognizing that combined exposures may pose greater health risks than individual pollutants. Predictive modeling incorporates historical exposure patterns, seasonal variations, and activity schedules to anticipate potential exposure events. Machine learning algorithms trained on occupational health databases continuously refine risk calculations based on emerging research and user feedback.Threshold Selection and Standards Compliance

[0152] The recommendation strength threshold against which the matching value is compared may be selected from multiple authoritative sources. These sources include, but are not limited to, occupational exposure limits established by regulatory agencies, absolute maximum concentrations for acute exposure scenarios, national guidelines such as EPA National Ambient Air Quality Standards, OSHA Permissible Exposure Limits, or HSE Workplace Exposure Limits, international standards including WHO Air Quality Guidelines or ISO workplace safety standards, recommendations from healthcare professionals including doctors, occupational health specialists, or care-providing professionals based on individual health conditions, and contextually-adjusted safety values customized for specific space types, materials, and activities. The system automatically selects the most appropriate threshold based on the detected pollutant, current activity context, user profile, applicable regulatory jurisdiction, and any medical recommendations. When multiple standards apply, the system may select the most protective threshold or provide graduated alert levels corresponding to different exposure criteria. Additionally, users may override automatic threshold selection and input custom threshold values through the user interface to accommodate specialized applications, experimental conditions, medical recommendations, or non-standard safety requirements specific to their workspace or operational needs.Dynamic Threshold Adjustment

[0153] The processor-readable instructions enable dynamic adjustment of recommendation strength thresholds based on real-time contextual information and user input. Static contextual information including space type, installed ventilation systems, and typical materials used establishes baseline threshold values during initial system configuration. Dynamic contextual information such as current activities, equipment operation status, and temporary conditions enables real-time threshold modification. For example, firing a kiln in a ceramics studio may temporarilylower particulate matter thresholds to account for increased silica exposure risk, while operating an exhaust fan may raise thresholds to reflect improved ventilation conditions. Machine learning algorithms analyze historical data patterns to optimize threshold values and reduce false positive alerts while maintaining appropriate safety margins.

[0154] In one embodiment, the system further comprises processor-readable instructions which train a machine learning algorithm using 9 degrees of freedom inertial measurement unit (IMU) data to classify normal operational environments of machines over a specified time period. The IMU provides comprehensive motion sensing through three-axis accelerometer data measuring linear acceleration, three-axis gyroscope data capturing rotational velocity, and three-axis magnetometer data detecting magnetic field orientation. This multi-sensor approach enables detailed characterization of machine vibration patterns, operational states, and anomalous behaviors.

[0155] The machine learning training process begins with a calibration period optimized for rapid deployment and user convenience. For basic operational state detection, the system can establish baseline signatures using a small number of vibrations recording samples for each operational state. Typically, users record multiple samples of machine operational states (e.g., "CNC cutting," "kiln heating," "planer running") and multiple samples of non-operational states (e.g., "machine idle," "powered off," "standby mode") for each piece of equipment. Sample durations are configurable based on machine characteristics and operational cycle times, enabling complete calibration within minutes for most workshop equipment. For more complex machines with multiple operational modes or variable cycles, extended calibration periods may be beneficial to capture operational variations such as different cutting speeds, material types, or thermal cycles. However, the system is designed to provide useful anomaly detection immediately after initial calibration, with accuracy improving as additional operational data is collected during normal use.

[0156] During calibration, the system collects baseline operational signatures for various equipment types including industrial machines such as CNC lathes, ceramic kilns, wood planers, and metal fabrication equipment, as well as residential appliances such as kitchen range hoods, HVAC systems, washing machines, dryers, and home workshop equipment including 3D printers and power tools. Users have flexibility to train and label operational data according to their specificworkshop contexts through simple mobile app interfaces that guide the sampling process. The system prompts users with instructions like "Start machine and record normal cutting operation" or "Record with machine powered off." Additionally, the system supports sharing anonymized baseline datasets between users operating similar equipment types, enabling new users to start with pre-trained models that can be refined with their specific machine samples. For instance, a typical operational environment for cutting medium-density fiberboard (MDF) produces distinctly different vibration signatures and particulate generation patterns compared to cutting acrylic materials, requiring separate trained models. Similarly, a kitchen range hood operating normally generates different IMU patterns than one with worn bearings or blocked filters, while simultaneously affecting air quality through cooking-related pollutant removal efficiency. The system implements multiple complementary threshold detection methodologies to accommodate varying technical requirements and computational capabilities across different deployment scenarios.

[0157] Amplitude-based threshold detection provides the most fundamental anomaly detection approach without requiring machine learning processing. This method establishes baseline amplitude ranges for each IMU axis during documented normal machine operation by collecting data samples at configurable intervals depending on machine characteristics. The system calculates statistical baselines (mean, standard deviation, minimum, maximum) for accelerometer and gyroscope readings across all three axes. Upper and lower threshold limits are then configured as percentages of baseline amplitude values, with typical settings ranging from ±15% for precision equipment to ±25% for heavy industrial machinery. Real-time monitoring continuously compares incoming IMU readings against these fixed threshold values.

[0158] Operational state classification occurs based on combined amplitude and temporal criteria, enabling correlation between machine behavior and environmental pollutant levels. The system tracks different operational states: brief amplitude exceedances (3-5 seconds) may indicate machine startup, load changes, or material transitions; moderate duration exceedances (10-30 seconds) suggest sustained operational modes such as heavy cutting, thermal cycling, or variable speed operation; and sustained threshold variations (60± seconds) indicate extended operational states such as continuous production runs, idle periods, or maintenance activities. Additionally, the system detects intermittent operational patterns through temporal analysis, where repeated brief threshold exceedances within specified time windows (e.g., five 3-second variations within a 10-minute period) characterize specific manufacturing processes, material handling operations, or cyclical equipment behavior. This combined amplitude-temporal classification enables precise correlation between machine operational states and corresponding environmental sensor measurements, allowing the system to associate specific pollutant generation patterns with particular equipment behaviors, material processing activities, or operational intensities. For example, the system can correlate sustained high-amplitude vibrations from a wood planer with increased particulate matter levels, or associate intermittent amplitude spikes from a CNC machine with volatile organic compound releases during different cutting operations. This approach requires minimal computational resources and no machine learning training, making it ideal for embedded systems with limited processing power or battery-operated portable monitors.

[0159] Frequency domain analysis employs Fast Fourier Transform (FFT) processing to extract spectral characteristics that reveal machine operational patterns invisible in time-domain analysis. The system performs FFT analysis on sliding windows of IMU data, with window sizes configurable depending on desired frequency resolution and computational constraints. During calibration, the system identifies dominant frequency components associated with normal machine operation, including fundamental frequencies, harmonic content, and spectral energy distribution across multiple frequency bands. These frequency bands can be configured to capture various operational characteristics including low-frequency structural vibrations, motor frequencies, cutting tool frequencies, bearing frequencies, and high-frequency tool chatter or wear signatures that may extend into kilohertz ranges for precision equipment. Baseline frequency signatures are characterized by peak frequencies, their amplitudes, harmonic ratios, and total energy within each frequency band. Anomaly detection occurs through multiple spectral criteria: frequency domain amplitude deviations exceeding configurable thresholds from baseline peak amplitudes, dominant frequency shifts beyond established tolerances from baseline frequencies, or spectral energy distribution changes exceeding predetermined variance limits within specific frequency bands. This approach effectively detects bearing wear, tool dulling, belt slippage, motor irregularities, and high- frequency tool condition changes.

[0160] Support Vector Machine (SVM) classification provides sophisticated pattern recognition for complex, multi-dimensional operational states that cannot be characterized by simple threshold methods. The SVM implementation processes temporal windows of IMU data ranging from 10-60seconds, with feature vectors comprising statistical measures calculated for each IMU axis within each window. Features typically include mean, variance, standard deviation, skewness, kurtosis, minimum, maximum, and range values for all nine IMU channels. Additional features may include cross-correlation coefficients between axes, spectral centroid, spectral roll-off, and zero-crossing rates. During training, the SVM algorithm establishes hyperplane decision boundaries in the highdimensional feature space that optimally separate normal operational states from various anomalous conditions. The system supports multiple kernel functions including linear kernels for linearly separable data, polynomial kernels for moderately complex patterns, and radial basis function (RBF) kernels for highly nonlinear operational signatures. Anomaly detection occurs when SVM confidence scores fall below configurable thresholds (typically ranging from 0.6-0.8 depending on desired sensitivity) or when the distance from current feature vectors to established hyperplanes exceeds trained margin values. The SVM approach excels at detecting subtle operational changes that might not trigger amplitude or frequency-based thresholds.

[0161] Least Squares Support Vector Machine (LSVM) regression extends the SVM approach by providing continuous prediction capabilities with error-based anomaly detection. Unlike classification-based SVM, LSVM learns to predict expected IMU values based on recent historical data and current operational context. The LSVM implementation solves regularized least squares optimization problems that map IMU feature inputs to predicted sensor outputs, effectively learning the normal relationship between machine state indicators and resulting vibration patterns. During calibration, prediction error baselines are established by analyzing residual error distributions from training data, calculating statistical parameters including mean prediction error, error variance, and error distribution shape. Real-time anomaly detection occurs when new IMU data predictions generate residuals exceeding 3 standard deviations (3o) from the established error distribution, indicating the current operational state differs significantly from learned normal patterns. The LSVM model supports adaptive learning through recursive least squares algorithms that update model parameters when new labeled operational data becomes available, enabling threshold adjustment as machines age or operating conditions change. This approach is particularly effective for detecting gradual performance degradation that might not trigger sudden threshold violations.

[0162] Neural Network Autoencoder implementation provides unsupervised anomaly detection by learning compressed representations of normal operational patterns. The autoencoder neuralnetwork consists of an encoder that compresses IMU feature vectors into lower-dimensional latent representations, and a decoder that reconstructs the original input from the compressed representation. During training on normal operational data, the autoencoder learns to minimize reconstruction error for typical machine behavior patterns. The network architecture can be configured with multiple hidden layers, with common configurations including input layers matching IMU feature dimensions, encoding layers that progressively reduce dimensionality, bottleneck layers representing compressed operational signatures, and decoding layers that mirror the encoder structure. Anomaly detection occurs when reconstruction error exceeds learned thresholds, indicating that current IMU patterns differ significantly from the normal operational patterns encoded during training. Reconstruction error thresholds are typically set at 2-3 standard deviations above mean training reconstruction error. The autoencoder approach is particularly effective for detecting novel anomaly types that were not present in training data, as it identifies deviations from learned normal behavior rather than requiring examples of all possible failure modes. Additionally, the compressed latent space representations can be analyzed to identify specific operational characteristics or failure patterns, providing interpretable insights into machine condition changes.

[0163] Temporal statistical analysis provides time-domain characterization complementing frequency-domain methods. Root mean square (RMS) vibration analysis calculates RMS amplitudes for each IMU axis over configurable time windows, typically ranging from 5-60 seconds depending on machine cycle times. RMS analysis effectively captures overall vibration energy levels and detects changes in machine dynamic behavior. Anomaly detection occurs when current RMS values exceed multipliers of baseline RMS values, with typical thresholds ranging from 1.3x to 2.0x baseline depending on machine type and desired sensitivity. Peak-to-peak amplitude analysis monitors maximum amplitude variations within time windows, capturing intermittent spikes or unusual excursions that might not affect RMS values. Deviation thresholds are typically configured at ±15% to ±30% from calibrated operational ranges. Pattern correlation analysis computes correlation coefficients between current IMU data segments and stored baseline operational patterns, providing a measure of similarity to known good operational signatures. Correlation-based anomaly detection triggers when correlation coefficients drop below configurable thresholds, typically ranging from 0.7-0.9 depending on pattern complexity and noise levels.

[0164] The classification system enables users to define custom operational categories tailored to their specific equipment and processes across industrial, commercial, and residential environments. Example categories include industrial applications such as "CNC machine cutting aluminum," "ceramic kiln firing cycle," "air filtration system normal operation," "wood planer surfacing hardwood," or "metal grinder sharpening tools," as well as residential applications such as "kitchen range hood normal operation," "HVAC system heating cycle," "washing machine spin cycle," "dryer normal operation," or "3D printer active printing." When machines operate within learned parameters, the system automatically timestamps database entries with trained classification outputs such as "CNC machine operating normally," "kiln temperature ramping," "filtration system anomaly detected," "range hood running efficiently," "HVAC compressor normal operation," or "washing machine unbalanced load detected." This automated labeling enables correlation analysis between machine operational states and environmental sensor data. When IMU data profiles deviate from trained classification models, the system flags these as anomaly events and prompts users to provide contextual information about unusual conditions, maintenance activities, or operational changes. User feedback helps refine classification accuracy and expand the system's knowledge of normal vs. anomalous operational patterns.

[0165] For multi-machine environments such as shared workshops, manufacturing facilities, or smart homes, the system supports networked monitoring of multiple IMU-equipped environment monitors. Each monitor maintains machine-specific trained models while contributing to shared pattern libraries that improve overall system performance. Anonymous data sharing enables users with similar equipment to benefit from collective operational knowledge without compromising proprietary information. The distributed approach allows workshop managers to monitor equipment health across multiple rooms or buildings through centralized dashboards showing real-time operational status, historical performance trends, and maintenance recommendations.

[0166] Integration between IMU-based machine monitoring and environmental air quality measurement provides comprehensive workspace safety monitoring across industrial, commercial, and residential applications. In manufacturing environments, when machine operational anomalies are detected through IMU analysis, the system simultaneously analyzes corresponding air quality measurements to determine environmental impact. For example, if a dust collection system attachedto a woodworking machine shows abnormal IMU patterns indicating potential filter blockage or fan problems, the system correlates this with particulate matter measurements to assess whether air quality is being compromised.

[0167] In residential applications, the system provides similar integrated monitoring for home appliances and systems. For instance, when an environment monitor attached to a kitchen range hood detects abnormal fan vibrations suggesting bearing wear or motor imbalance, the system correlates this with air quality measurements to determine if cooking-related pollutants (particulate matter from frying, volatile organic compounds from sauteing, nitrogen dioxide from gas burners) are being effectively removed. The system can alert users with messages such as "Range hood running but PM2.5 not decreasing - check filter or ductwork." Similarly, for HVAC systems, the monitor detects normal operational vibrations versus problems such as belt slippage, bearing wear, or compressor issues, while simultaneously tracking filter effectiveness through particulate reduction monitoring, detecting potential refrigerant leaks, and monitoring humidity control performance during heating and cooling cycles. In laundry applications, IMU sensors detect normal wash / spin cycles versus unbalanced loads or mechanical problems, while environmental monitoring includes lint detection for dryer fire prevention, humidity tracking for proper ventilation, and volatile organic compound detection from fabric softeners or cleaning products.

[0168] This integrated approach enables proactive maintenance scheduling that addresses both equipment reliability and environmental health protection across diverse applications. Users receive recommendations combining machine maintenance actions with environmental safety measures, such as "Replace dust collector filter and increase ventilation," "Service CNC spindle bearings and monitor VOC levels during operation," "Clean range hood filter and check exhaust ductwork," or "Balance washing machine load and improve laundry room ventilation."

[0169] In one embodiment, the system incorporates a specialized large language model (LLM) comprising a transformer-based neural network architecture specifically fine-tuned for environmental health and safety applications. The LLM architecture is designed for flexible deployment, supporting both cloud-based implementation for comprehensive analysis and edgedevice deployment for reduced-parameter models optimized for real-time audio feedback and offline functionality. The system utilizes task-specific model optimization to reduce computationalrequirements while maintaining domain expertise for air quality and occupational health applications.

[0170] The large language model (LLM) is trained on a comprehensive and domain-specific corpus of datasets curated to support its application in environmental health and safety. These datasets include regulatory and advisory documents from authoritative bodies such as the Occupational Safety and Health Administration (OSHA), covering workplace safety regulations and guidelines; and the Environmental Protection Agency (EPA), encompassing air quality standards and public health advisories. Additionally, the LLM leverages international health and safety standards such as ISO 45001 and ANSI Z87.1 to ensure global applicability.

[0171] The training corpus further includes Material Safety Data Sheets (MSDS) from both industrial and creative material databases, enabling material-specific hazard recognition. Risk profiles of various wood species and material-specific health databases provide insights into potential respiratory and carcinogenic hazards. The model also incorporates Health and Safety Executive (HSE) documentation focused on pollutant risk assessment, including detailed data on silica exposure.

[0172] To support comprehensive exposure limit evaluations and risk assessments, the LLM is trained on authoritative workplace exposure documentation, including EH40 / 2005 Workplace Exposure Limits published by the Health and Safety Executive (HSE), Threshold Limit Values (TLVs) from the American Conference of Governmental Industrial Hygienists (ACGIH), and international exposure standards from regulatory bodies including EPA, OSHA, NIOSH, WHO, and equivalent international health organizations. The training corpus incorporates peer-reviewed scientific literature on occupational health risks within creative and industrial settings, toxicological databases, carcinogenicity assessments from the International Agency for Research on Cancer (IARC), and material-specific health impact studies to enhance the model's evidence-based reasoning capabilities. The LLM training dataset further encompasses equipment operation manuals and safety protocols for commonly used machinery in workshops, including but not limited to woodworking equipment, metalworking tools, ceramics kilns, 3D printers, laser cutters, and air filtration systems. Real-world case studies and incident reports sourced from creative workspaces, industrial hygienist assessment reports, workplace injury databases, and emergency responsedocumentation provide practical context for safety recommendations. Additionally, the model incorporates Material Safety Data Sheets (MSDS) from industrial and creative material databases, building science research on pollutant dispersion and ventilation effectiveness, and sensor calibration studies to ensure accurate interpretation of environmental monitoring data and generation of contextually appropriate safety guidance.

[0173] International resources such as the World Health Organization (WHO) air quality guidelines, ISO environmental and safety standards, and green building certification frameworks, including RESET, WELL, LEED, and Green Globes, contribute additional layers of safety reference. The LLM also draws from healthcare advisories provided by physicians, care providers, and medical institutions, allowing for integration of clinical insights into recommendation logic.

[0174] Geographical diversity is ensured by incorporating regulatory frameworks from a broad set of regions including the United States, United Kingdom, European Union, Southeast Asia, Korea, Japan, Australia, India, Mexico, Brazil, Thailand, Singapore, China, and the United Arab Emirates, thereby enabling context-aware recommendations tailored to local regulations and practices.

[0175] Furthermore, the LLM is trained to understand and correlate complex relationships between various contextual factors. These include material properties and their associated health risks, such as the carcinogenicity of hardwood dust versus softwood, or the risk of silicosis due to silica exposure; activity-specific exposure patterns and the corresponding protective measures; the influence of environmental conditions on pollutant dispersion and behavior; temporal aspects such as exposure duration, frequency, and cumulative effects; and individual user profiles that take into account demographic data and underlying health conditions. This contextual knowledge integration enhances the system’s ability to deliver personalized, relevant, and evidence -based environmental health and safety recommendations. The LLM system is configured to process and analyze contextual information from multiple input modalities through specialized preprocessing modules.

[0176] A text analysis module Processes user-inputted descriptions of workspace activities, materials, and equipment using named entity recognition (NER) and semantic parsing to extract relevant safety parameters. A voice processing module Incorporates automatic speech recognition(ASR) with domain-specific vocabulary for workshop terminology, converting spoken descriptions into structured contextual data. Image analysis integration interfaces with computer vision models to process visual contextual information, correlating detected objects, materials, and workspace configurations with the LLM's knowledge base. Sensor Data Interpretation analyzes patterns in environmental sensor readings in conjunction with contextual information to identify anomalous or elevated ambient pollution conditions and potential health risks

[0177] Multiple pollution patterns that frequently occur in correlation with specific activities can be identified by an Al system, even when such patterns might otherwise be overlooked. For example, 3D printer failure detection can be performed using air pollution data. In such cases, occupants may be advised on methods to avoid and stop print failures, receive purification and ventilation alerts, and be guided through automating integration with air purifiers. The system can also be configured to set up customer alerts before wood dust extraction filters become full, helping ensure that occupants are not frequently exposed to pollution due to unawareness of filter saturation. Additionally, anomaly detection is employed to alert users when new pollutant patterns are observed that do not typically appear in the established weekly pollution profiles, which may include momentary peaks that exceed recommended levels.

[0178] Advantageously, the system and method for environmental monitoring is adaptable, customizable, and provide more targeted solutions for improving air quality. The system incorporates contextual information about the type of workspace or creative space, the types of activities conducted, and the equipment used. The system offers flexibility in selecting sensors and provide specific, actionable advice tailored to individual environments. The system integrates seamlessly with other smart home devices for automated air quality management. Furthermore, the system allows users to select sensors based on the materials they work with, the processes they apply, and the pollutants of concern within their workspace. Additionally, the system includes sensors with air filters (such as HEPA filter and / or carbon filters) capable of detecting and mitigating harmful pollutants like particulate matter, volatile organic compounds (VOCs) and carbon dioxide (CO2) automatically.

[0179] While the disclosure has been described with reference to exemplary embodiments, it will be understood by those skilled in the art that various changes may be made and equivalentsmay be substituted for elements thereof without departing from the scope of the disclosure. In addition, many modifications may be made to adapt a particular system, device, or component thereof to the teachings of the disclosure without departing from the essential scope thereof. Therefore, it is intended that the disclosure not be limited to the particular embodiments disclosed for carrying out this disclosure, but that the disclosure will include all embodiments falling within the scope of the appended claims. Moreover, the use of the terms first, second, etc. do not denote any order or importance, but rather the terms first, second, etc. are used to distinguish one element from another.

[0180] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the disclosure. As used herein, the singular forms “a,” “an” and “the” are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms “comprises” and / or “comprising,” when used in this specification, specify the presence of stated features, integers, steps, operations, elements, and / or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0181] The description of the present disclosure has been presented for purposes of illustration and description, but is not intended to be exhaustive or limited to the disclosure in the form disclosed. Many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope of the disclosure. The described embodiments were chosen and described in order to best explain the principles of the disclosure and the practical application, and to enable others of ordinary skill in the art to understand the disclosure for various embodiments with various modifications as are suited to the particular use contemplated.

Claims

AMENDED CLAIMS received by the International Bureau on 14 November 2025 (14.11.2025)What is claimed is:1.A system for monitoring air quality and providing recommendations for managing air quality, comprising: at least one environment monitor comprising one or more environment monitoring sensors, wherein the environment monitor is configured to sense environmental data in a space, wherein the environment monitoring sensor comprising at least one of a formaldehyde sensor, a carbon monoxide sensor, a nitrogen oxides sensor, an ozone sensor, a radon sensor, a passive infrared sensor, an ultraviolet light sensor, and an inertial measurement unit; a user device associated with a user, wherein the user device is configured to receive user input comprising contextual information about the space; a database configured to store the environmental data and the contextual information, and a server in communication with the database, the user device and the environment monitor, wherein the server comprises one or more artificial intelligence modules, wherein the server is configured to analyse the contextual information and environmental data, and provide recommendations for managing air quality in the space, wherein the user device is configured to enable the user to input contextual information to specify the space and the use of the space in which the environment monitor is placed, wherein the contextual information comprising: information selected from a floor area of the space; a height of the space; a volume of the space; a type of product stored in the space; a type of tool in the space; a type of machine in the space; a type of activity for which the space is typically used; a type of activity occurring in the space; a type of material processed in the space; a type of furniture in the space; a type of ventilation system available in the space; a type of air filtration system available in the space; a geographic location of the space; a type of building in which the space is located a type of heating and cooling system available in the space;and sources of pollutants within the space; a user’s sensitivity to pollutants; a user’s respiratory health condition.2.The system of claim 1 , wherein the environment monitoring sensor further comprising at least one of a volatile organic compound sensor, a particulate matter sensor, a carbon dioxide sensor, a temperature sensor, a relative humidity sensor, a radiation sensor, auditory sensor, and an infrared thermal sensor.3.The system of claims 1 to 2, further comprising: processor-readable instructions which, when executed by one or more processors, cause the one or more processors to: generate, via at least one of the machine interfaces, a recommendation to adjust alert profiles associated with the contextual information; provide, via a user input to the machine interface, acceptance of the recommendation to adjust alert profiles associated with the contextual information; cause execution, at the environmental monitor, of an instruction to set the adjusted alert profile associated with the contextual information based on the acceptance of the recommendation received via the machine interface.4.The system of claims 1 to 3, further comprising: processor-readable instructions which, when executed by one or more processors, cause the one or more processors to: store and / or retrieve data to / from a database, comprising data from at least one composition selected from: the environment monitoring sensor, a processor-generated timestamp, and the contextual information.

5. The system of claim 4, wherein a user input format for the contextual information comprising at least one of audio, image, video, text, GPS location, wireless signals, Bluetooth, Wi-Fi status, data from lidar sensor, data from depth sensing sensor, data from the environmental monitor, data from a networked multiple environment monitor, and data from networked devices.6.The system of claims 4 and 5, wherein the contextual information extracted via a machine learning algorithm, features from an input file.7.The system of claim 6, wherein the input file comprises, at least one of an photo of the space, a video, a live camera feed, and the machine learning algorithm extracts contextual information from the scene, comprising, including but not limited to, materials in the space, products placed in the space, such as cleaning supplies, electronics, estimated dimensions of the space, volumetric size of the space, equipment in the space, ventilation and air filtration system in the space, heating system for the space, type of windows, presence of dehumidifiers in the space, dimensions of windows, presence of plants, quantity of plants with air purifying capability, type of flooring, placements within the space of each object and whether the space is shared.8.The system of claims 6 and 7, further comprises: processor-readable instructions which, when executed by one or more processors, cause the one or more processors to: analyse live and historical data collected from the environment monitor; determining, from user input, a contextual information; based on determining the contextual information: analyzing, by a processor, a user profile for the user and determining one or more categories from the user profile that are related to the contextual information and / or data stored in the database; determining whether to trigger an item recommendation by comparing the matching value to a recommendation strength threshold; and when the matching value exceeds the strength threshold: triggering the one or more candidate item recommendations, and providing, recommendations to mitigate environmental health risks associated with the contextual information, wherein the recommendation strength threshold is determined based on at least one of: a predetermined ratio of regulatory exposure limits, contextual workspace risk factors, and user-configurable safety margins, wherein the matching value is a measured pollutant concentration, to a recommendation strength threshold selected from occupational exposure limits, absolute maximumconcentrations, national and international guidelines and standards, or contextually-adjusted safety values based on the space type, material, and likelihood of pollutant being present in the air.9.The system of claims 2 to 8, wherein the user device is a computing device, configured to utilize artificial intelligence algorithm and a natural language interface to generate recommendations for a user.

10. The system of claims 2 to 9, wherein the user device is configured to inform users of additional or relevant sensors to install and / or replace, and enable to expand on the visibility of the status of the environment data, display, large screen, physical display.

11. The system of claim 1, further comprising processor-readable instructions which, when executed by one or more processors, cause the one or more processors to: detect movement of the environmental monitor, generate a notification in response to the detected movement; label the timestamped data with movement information; and / or prompt the user to input an ongoing activity in the space.

12. The system of claim 1, wherein the multiple environment sensors can be placed to collect data within different spatial positions and / or rooms within a building to train a machine learning algorithm to infer how air pollutants vary, and estimate levels of pollution in the second room without the second sensor, wherein the system is trained to understand contextual activities and how it influences spatial pollution distribution, wherein the events such as opening a window in the second room, leading to improved air quality in the primary room, can be labeled by the user, thereby leading to an additional options for improving air quality in a space through the enhanced spatial pollution datasets.

13. The system of claim 1, further comprising processor-readable instructions which, when executed by one or more processors, cause the one or more processors to: train a machine learning algorithm with a set of air pollution data over a specified time period, to classify a normal operational environment of the space, wherein the users can train and label the data on their own, or share the dataset with others, for example, a typical operational pollutionenvironment for cutting MDF, will differ from cutting acrylic, wherein the classification machine learning algorithm then allows pollution profiles that do not fit its classified dataset to trigger an alert, and prompt users to label the activity and / or evaluate the context.

14. The system of claim 1, further comprising processor-readable instructions which, when executed by one or more processors, cause the oonnee or more processors to: train a machine learning algorithm with a set of air pollution data and / or IMU vibration data over a specified time period, to classify a normal operational environment of the space, wherein the users can train and label the data on their own, or share the dataset with others, for example, a typical operational pollution environment for cutting MDF, will differ from cutting acrylic, and corresponding IMU vibration signatures will differ between normal equipment operation and equipment anomalies, wherein the classification machine learning algorithm then allows pollution profiles and / or vibration patterns that do not fit its classified dataset to trigger an alert, and prompt users to label the event and / or evaluate the context; analyze cumulative contextual data over time, including but not limited to the environmental data, and / or IMU vibration data, and the activity data, to provide recommendations to improve air quality and / or predict when to automatically operate air purifiers, wherein the system automatically programs purifiers or HVAC systems to operate predictively to clean the space based on learned pollution patterns, and / or equipment vibration signatures, and anticipated activities, wherein the predictive control system optimizes equipment operation timing to prevent pollution accumulation rather than responding reactively to threshold violations.

15. The system of claim 1, wherein the user device is configured to: capture contextual information via one or more input formats including audio, image, video, text, GPS data, wireless signal data, Bluetooth status, Wi-Fi status, lidar sensor data, or depth sensor data, wherein the server is configured to extract the contextual information using a machine learning model, wherein the user device comprising at least one of an inertial measurement unit (IMU), a camera, or a depth sensor, wherein the user device is configured to map volumetric dimensions of a space and generate a three-dimensional spatial model of the space based on sensor data, wherein the user device communicates the three-dimensional spatial model as contextual information to the server, andreceive contextual information related to a temporary health condition of the user, wherein the contextual information including health condition is stored in the database with a timestamp and correlated with the environmental data collected by the environment monitor to identify pollutant exposure patterns, and provide personalized air quality recommendations to the user via the user device based on analysis of the correlated environmental data, contextual information, health condition patterns, and the three-dimensional spatial model, wherein the recommendations are tailored to the specific space geometry and user health sensitivity profile.

16. The system of claim 1, wherein the server is configured to: detect presence data of a user in the monitored space using at least one of: a presence sensor, thermal imaging sensor, camera, and GPS location, Wi-Fi status, or Bluetooth activity of the user device, wherein the presence data is timestamped and stored in the database; compute a time-weighted average (TWA) of pollution exposure for the user based on proximity to the environment monitor, display the calculated TWA exposure through the user device and issue alerts if the TWA exceeds predefined thresholds; prompt the user to input specific contextual information in response to a detected activity in the monitored space via the user device, prompt the user to add a new sensor module based on contextual information indicating a risk not covered by existing sensors, and generate and provide health risk recommendations based on the detected presence of hazardous substances and associated pollutant levels in the environment, and presence data of the user, wherein one or more environment monitors are networked to enable spatialanalysis across multiple rooms, and machine learning models are used to estimate pollution distribution across unmonitored regions based on contextual information.

17. A method for monitoring air quality and providing recommendations for managing air quality, comprising the steps of: inputting contextual information about the space in which the environment monitor is placed; providing informational advice to the user about risks associated with this particular space based on the contextual information; and prompting the user to change features and content displayed on a machine interface comprising of at least one of a mobile device, computer, and interface on the environment monitor, based on the contextual information, wherein the recommendations is generated from the contextual information and environmental data collected from the environment monitor.

18. A computer-implemented method for monitoring air quality and providing recommendation for managing air quality, comprising: receiving real-time environmental sensor data from at least one environment monitoring sensor; receiving contextual information about workspace activities, materials, and equipment; processing the environmental sensor data and contextual information through a large language model trained on occupational health and safety datasets; correlating material-specific health risk profiles with detected environmental conditions; generating contextually relevant health and safety recommendations that account for both immediate exposure conditions and cumulative health risks; providing the recommendations through a user interface in a format adapted to user expertise level; collecting environmental sensor data; analyzing daily and weekly pollution patterns to establish baseline conditions; correlating pollution patterns with user-provided contextual information about activities and materials;integrating external environmental data including weather patterns and local air quality measurements; generating workspace-specific pollution fingerprints that distinguish normal operational patterns from anomalous conditions; providing anomaly alerts when detected pollution patterns deviate from established fingerprints; monitoring environmental conditions during sequential workspace activities; identifying activity combinations that create enhanced health risks beyond individual activity exposure; analyzing chemical interaction effects from multiple materials or processes; generating recommendations for activity sequencing and intermediate protective measures; providing compound exposure warnings when activity combinations exceed safe exposure thresholds; collecting cumulative environmental exposure data over extended time periods; receiving user health symptom reports and health status updates; correlating exposure patterns with reported health symptoms using machine learning algorithms; developing personalized health risk profiles based on individual exposure sensitivity, and generating predictive health risk assessments and preventive intervention recommendations.

Images