Voice Identification

Authentication and identification are critical for the acceptance of many applications, at home and in office environments. While for example face recognition is applied since years, authentication or even identification based on the speaker’s voice is still an evolving topic. Objective of this project is to develop a system for voice identification based on short sequences of speech, e.g. 1 minute for training and 10 seconds for recall, with free text, robust against a certain degree of background noise, and suitable for execution on an embedded processor. For the training much more computing power can be used, incl. GPUs. Algorithm development can be performed on Windows or Linux, involving signal processing frameworks such as Matlab and neural network frameworks such as PyTorch. The voice identification will be integrated into a complete HMI system. Knowledge in C/C++, Python is surely helpful. Knowledge in signal processing is required. Usually students start in our department with an internship or as working student, and then have the chance to continue their work within a master thesis, if supervision by their university is provided. Start: As soon as possible, but not before November 15th, 2018 (1st and 15th day of each month are possible entry dates). Location: Infineon Campeon, Neubiberg