Some Trials of Blind Source Separation
We show some results of our experiments on Blind Source Separation. In every experiment, the number of the inputs and the outputs is 2. The following files are all stereo sound files. Left channel and right channel correspond to the inputs to the microphne 1 and microphne 2, respectively. Links to our matlab codes and C codes (last updated in 1998).
N. Murata, S. Ikeda, A. Ziehe, "An approach to blind source separation based on temporal structure of speech signals,'' Neurocomputing, 41(1-4), 1-24, Oct. 2001 |
---|
S. Ikeda, N. Murata, "A method of blind separation based on temporal structure of signals,'' ICONIP'98, 737-742, Kitakyushu, Japan, Oct. 1998 |
N. Murata, S. Ikeda, "An on-line algorithm for blind source separation on speech signals,'' NOLTA'98, 923-926, Crans-Montana, Switzerland, Sep. 1998 |
I. Signals Mixed on Computer
II. Signals from Dr. Te-Won Lee's Home Page
III. Signals from Mr. Paris Smaragdis' Home Page
IV. Signals from Prof. Kota Takahashi
V. Signals from Dr. Mark Girolami
VI. Online Learning Algorithm
I. Signals Mixed on Computer
These examples are artificially mixed on DEC Alpha station 200 after recording each source signal separately. For the detail, please go to the description page.
1. Instantaneous mixtures
- Mixed Sounds (stereo)
- Our Results
- Separated Source 1 (stereo)
- Separated Source 2 (stereo)
2. Convolutive mixtures I
- Mixed Sounds (stereo)
- Our Results
- Separated Source 1 (stereo)
- Separated Source 2 (stereo)
3. Convolutive mixtures II (in a virtual room)
- Mixed Sounds (stereo)
- Our Results
- Separated Source 1 (stereo)
- Separated Source 2 (stereo)
II. Signals from Dr. Te-Won Lee's Home Page
These examples are taken from Dr. Te-Won Lee's Home page at the Salk Institute.
1. Speech - Music Separation
- Recorded Sound (stereo)
- Our Results
- Separated Source 1 (stereo)
- Separated Source 2 (stereo)
2. Speech - Speech Separation
- Recorded Sounds (stereo)
- Our Results
- Separated Source 1 (stereo)
- Separated Source 2 (stereo)
3. Speech - Speech Separation in difficult environments
- Recorded Sounds (stereo)
- Our Results
- Separated Source 1 (stereo)
- Separated Source 2 (stereo)
III. Signals from Mr. Paris Smaragdis' Home Page
This example is taken from Dr. Paris Smaragdis' Home Page in 90's.
- Input Sounds
- Microphone 1 (monoral)
- Microphone 2 (monoral)
Sampling Rate was modified for our experiment (Input stereo sound). We need the data to be sampled with 16 kHz. This is because of the practical reason of the calculation. We are using MATLAB, and restricted amount of memory. So, we changed the sampling rate from 44.1 kHz to 16 kHz with free software "sox".
- Our Results
- Separated Source 1 (stereo)
- Separated Source 2 (stereo)
IV. Signals from Prof. Kota Takahashi
These data were given by Prof. Kota Takahashi (Japanese only) in the University of Electro-Communication.
1. Speech - Speech Separation in difficult environments
These data were recorded in a real environment.
- Input sound (stereo)
- Our Results
- Separated Source 1 (stereo)
- Separated Source 2 (stereo)
2. Speech - Speech Separation in difficult environments
This data were also recorded in the same lab. But they include more reflections.
- Input sound (stereo)
- Our Results
- Separated Source 1 (stereo)
- Separated Source 2 (stereo)
V. Signals from Dr. Mark Girolami
Dr. Mark Girolami gave me some data, and I used them for experiments.
1. Speech - Speech
These data were recorded in a real environment.
- Input sound (stereo)
- Our Results
- Separated Source 1 (stereo)
- Separated Source 2 (stereo)
2. Speech - Noise (Gaussian)
- Input sound (stereo)
- Our Results
- Separated Source 1 (stereo)
- Separated Source 2 (stereo)
3. Speech - Noise (Laplace)
- Input sound (stereo)
- Our Results
- Separated Source 1 (stereo)
- Separated Source 2 (stereo)
4. Speech - Noise (Gaussian)
- Input sound (stereo)
- Our Results
- Separated Source 1 (stereo)
- Separated Source 2 (stereo)
5. Speech - Noise (Laplace)
- Input sound (stereo)
- Our Results
- Separated Source 1 (stereo)
- Separated Source 2 (stereo)
VI. Online Learning Algorithm
These examples are artificially mixed on DEC Alpha station 200 after recording each source signal separately.
1. Online Learning
- Mixed Sounds (stereo)
- Our Results
- Separated Source 1 (stereo)
- Separated Source 2 (stereo)
Belief Propagation
S. Ikeda, T. Tanaka, S. Amari, "Stochastic reasoning, free energy, and information geometry,'' Neural Computation, 16(9), 1779-1810, Sep. 2004 |
---|
S.keda, T. Tanaka, S. Amari, "Information geometry of turbo and low-density parity-check codes,'' IEEE transaction on Information Theory, 50(6), 1097-1114, Jun. 2004 |
Acceleration of the EM algoritum
S. Ikeda, "Acceleration of the EM algorithm,'' Systems and Computers in Japan, John Wiley & Sons, Inc., 31(2), 10-18, Feb, 2000 |
---|
Channel Capacity
S. Ikeda, J.H. Manton, "Capacity of a single spiking neuron channel," Neural Computation, 21 (6), 1714-1748, 2009 |
---|
S. Ikeda, K. Hayashi, T. Tanaka, "Capacity and modulations with peak power constraint,'' arXiv:1005.3889 |
Rate-Distortion Function
K. Watanabe, S. Ikeda, "Rate-distortion function for gamma sources under absolute-log distortion measure," ISIT2013, 2557-2561, Istanbul, Turkey, Jul. 2013 |
---|
X-ray diffraction
S. Ikeda, H. Kono, "Phase retrieval from single biomolecule diffraction pattern," Optics Express, 20(4), 3375-3387, 2012 |
---|
Interferometer
M. Honma, K. Akiyama, M. Uemura, S. Ikeda, "Super-resolution imaging with radio interferometer using sparse modeling," arXiv:1407.2422 |
---|
Compton Camera Imaging
S. Ikeda, H. Odaka, M. Uemura, T. Takahashi, S. Watanabe, S. Takeda, "Bin Mode Estimation Methods for Compton Camera Imaging," arXiv:1312.4291 |
---|