Higher Order Conditional Random Fields Aligner (HOCRFAlign): [matlab_tar.gz]
Input: query protein sequence and PDB file of template protein. Output: estimated PDB file of query protein
Speech
Uncertainty propagation on HMM+DNN: [matlab_tar.gz]
Input: featuregram and uncertaintygram. Output: HmmLikelygram based on Monte-Carlo, Entire,.... uncertainty propagations.
Speaker Localization (SpLoc) and Beamforming (BForm): [matlab_tar.gz]
SpLoc based on Position and Pitch (PoPi) plane and BForm based on Delay and Sum (DS).
Deep Belief Network Voice Activity Detector (DBN-VAD) and Room Localizer (RLoc): [matlab_tar.gz]
From signal to DBN-VAD and RLoc. Training feature extraction, training and test are possible.
Create embedded noisy (EmbNoisy) signals using dry (non-reverberated) clean isolated (DryIsoClean) signals, impulse responses (IRs) and noises (Noises): [matlab_tar.gz]
From DryIsoClean to EmbNoisy folder at a specific SNR. For DryIsoClean, IRs and Noises ask DIRHA or FBK members.
Vector Taylor Series Compensation (VTSComp): [matlab_tar.gz]
VTSComp in filterbank (log-Mel) domain using a Gaussian Mixture Model (GMM). We generate a fake speech database to do training feature extraction, training and test.
Some feature extractors used in my papers: [matlab_tar.gz]
Assymetric Window, Sifting Autocorrelation, Tunnelling Noise, DTW pitch, etc..
Basic Feature Extractor (FE): [matlab_tar.gz]
From signal to segmentation, spectrogram, cepstrogram and cepstrogram+deltas.
Real-time DSR (Distributed Speech Recognizer) over IP (TCP, UDP, RTP) [c_tar.gz]
Linux and Windows C sources. Word project.
Electronics and Robotics
Greenhouse web (IoT) . In Python, PHP and Arduino.
Input: actuators wished values (roof angle, watering seconds, etc.). Output: system state (temperature, humidity, etc.).
