Data Reduction Techniques for Brain Machine Interfaces [MP, bt/mt]

Universität Ulm

M. Pagin

One of the greatest challenges that arises in these systems is transmission of the recorded data from the brain. A 100 channel recorder can easily generate tens of Mbit/s of data rate. With the power constrains given for biomedical implants this data rate exceed the maximum that is allowed to be transmitted. It is necessary to implement ad-hoc solutions for compressing the data.

Compression needs to be effective in reducing the data rate and at the same time don’t affect signal elaboration, known as spike sorting process. This process reconstructs which neurons are active at a time in the brain and classifies them, clustering.

We offer thesis on a broad variety of topics:

  • Evaluation and design of signal dependent ADCs
    These ADCs adapts their resolution to the incoming signal, using higher resolution only in the useful parts of the signal.
  • Digital compression schemes
    Predictor and entropy reduction systems, combine a predictor and an entropy encoder to achieve lossy and lossless compression. For example delta compression and Huffman encoding.
  • Compressed sensing schemes
    Exploit signal sparity and structure in order to reduce the number of samples needed to acquire the signal below Nyquist rate. 

Theses involve evaluation of the scheme (Matlab/Simulink) and analog or digital implementation (Cadence and VHDL/Verilog) depending on the topic.

What we expect: Basic knowledge of electronics. Basic know how of Matlab/Simulink and Cadence (for ADCs topic only). Organized and well documented research and dedication to successful work.

Zurück zu Stellenangebote