Project: Auditory Scene Analysis
The Philips Hue system and products (www.meethue.com) are constantly evolving and improving the way users interact with light at their homes. Within Philips Lighting Research we are currently investigating new ways for the Hue system to be smarter and understand what is happening to the user and its surroundings. One of this mechanisms is by means of sound, which can enable great possibilities for our customers.
This assignment consists of evaluating the feasibility of audio processing for feature extraction in homes, taking into account performance of multiple algorithms, impact of the extracted data, and portability of SW to embedded platforms. This will be carried out as part of a larger team of scientists that will also evaluate use cases, HW requirements, and overall applicability of this technology.
What is expected from a student is:
Investigate from literature on known algorithm libraries, as a function of predefined user features
Define key performance indicators (KPI) together with system architect
Compare selected algorithms based on KPI, and select highest scoring solution
Generate PC-based test environment (e.g. Matlab) for algorithm development and testing
Validate algorithm based on collected data (generated by project team)
Port algorithm to embedded platform for further project testing (e.g. Raspberry PI, Arduino, microcontroller).
Communication of results and trade-offs to project team and stakeholders
Expected student profile for this assignment is as follows:
Strong signal processing skills
Knowledge of audio-based phenomena (preferred)
Skilled in scientific/algorithm programming languages (e.g. Matlab, etc.).
Knowledgeable in embedded SW (C, C++, Linux, RTOS) (preferred)
For further information, please contact J.P.M.G. Linnartz (email: J.P.Linnartz@) tue.nl