Voice Recognition for ID&V
The client, a small and medium retail bank, was facing a lot of customer dissatisfaction with respect to the telephone banking authentication method. The existing system wanted the customer to remember a combination of digits, dates etc. in order verify their identity and this process could take up to 3-5 minutes if the customer forgets these pin numbers and dates. In order to reduce the identity and verification on the telephone banking, the client wanted to have a voice recognition system to verify the customer’s identity.
Python, Keras, TensorFlow, SQL (for extracting voice properties)
We proposed a voice recognition system to be integrated with the existing the telephone banking. The idea was to store the customer voice (by saying pre-defined sentence) in the bank’s data center and each of these biometric voice snippet would be unique for each caller. Now, every time the customer calls for telephone banking, his voice will be automatically compared with the saved voice print in the data center with the help of a back end machine learning model. This reduced the verification process time to under 30 seconds and the bank received a high customer satisfaction rating post this implementation.
Call: 020 3573 6917