Final project for the Building AI course
The concept revolves around identifying distinctive features within the sound profile of a specific vehicle type and utilizing emerging deviations from this type as warning signals.
Vehicle issues, particularly those related to engines and transmissions, are common occurrences and can lead to costly repairs. My motivation stems from the desire to provide a solution that contributes to the early detection of problems, thereby preventing more severe and expensive damage. Additionally, the diagnostic process is likely to be cost-effective as well, compared to inspections that may require complex dismantling work. Moreover, considering that on-site inspections may not even be feasible in certain areas, this approach has the potential to address issues in regions with limited vehicle maintenance infrastructure, making it indispensable.
The solution enables fast and precise diagnosis of vehicle issues by recording engine noises using a handheld device. The application analyzes the recorded sound, identifies the vehicle make and model, and conducts a comprehensive engine diagnosis. The output includes information about identified problems and their urgency. With an internet connection, the application accesses an extensive database. In offline mode, it relies on predefined diagnosis data.
Data sources include sound files from vehicles, engines, or transmissions of a specific type. Quality and availability of sound data are essential. I will employ AI techniques for pattern recognition and analyzing deviations in sound profiles. Application: The solution will be employed in vehicle maintenance and repair, benefiting vehicle owners and fleet operators. Thanks to its suitability for regions with limited infrastructure for vehicle analysis, such as emergency areas and developing countries, it serves not only commercial but also humanitarian purposes by enhancing vehicle safety and reliability.
The current concept's weakness lies in internet connectivity, especially in underdeveloped regions and humanitarian applications. The challenge is to develop a practical and conceptual solution that optimally utilizes the handheld device's storage capacity in regional contexts and minimizes dependency on the internet.
In the future, I could expand the project to cover other vehicle types or specific diagnostic functionalities, offering a more comprehensive solution for vehicle maintenance and repair.
The basic idea for our project was inspired by Bavarian inventor Michael Schmutzenhofer who successfully detects odometer manipulations through sound analysis (though using ultrasound). Proper credit will be given to all utilized sources and resources, especially if accessing open-source data or code.