Lloyds Hackathon Cancara

My idea was to create a virtual assistant which can understand conversational English to process a financial request. The user should be able to press a physical button, have some sort of verification to ensure the person using the device has access to the account, and then start speaking to the device as if it was a person in a Lloyd’s bank branch. The device should be able to respond to the user through speech.

Firstly, when run for the demonstration a window will appear. This window contains a basic design of the device, consisting of a circular button with the word “Cancara” written across it (created in pygame). This window acts as the device would in real life. When the button is clicked, a web camera live stream is created. This live stream uses OpenCV to detect and recognise faces. This facial recognition is used for security. Once the user’s face has been recognised and linked to their account, Cancara will start listening to the user through their default microphone. Once the user has finished speaking, Cancara will process the request and either complete the task or ask necessary follow-up questions. Currently, Cancara can add payment contacts and make fictional money transfers. Since this was a proof-of-concept project, it did not need to provide financial advice. I would like to improve on this project in the future to provide such information.

Not only does this replace the need for a branch, but using this system is much quicker than using an app. Instead of clicking through multiple screens and buttons, you can simply say “transfer this person this amount of money” and the transfer is made. This would be especially useful if connected with other popular virtual assistants, such as Siri.