Musical Instrument Sound Separation: Extracting Instruments from Musical Performances - Theory and Algorithms
In this book you will find three algorithms for separating the audio waveform of instruments in musical performances. Each of these algorithms uses stereo cues and makes some assumptions about the structure of music in order to exploit the resulting redundancy of the tones and to localize the instruments in space. Much detail is given on the development of the algorithms, ranging from the idea at the beginning with the initial assumptions to the final algorithm. Suggestions for fine tuning are presented along with extension ideas for each algorithm in part. A separate chapter is focusing on implementation details showing ways to get the maximum speed out of the available hardware and software. A review of the state of the art algorithms in scientific literature is also included along with the common problems encountered. The lack of tools to consistently evaluate the separation quality is considered and consequently two subjective quality criteria are introduced together with a new testing corpus."
What people are saying - Write a review
We haven't found any reviews in the usual places.
Direct Template Matching
Iterative Template Matching
7 other sections not shown
assigned Auditory Scene Analysis auditory system baseline BASS-dB binaural cues binaural recorded blind source separation calculate channel Chapter cluster colour computational convergence convolution corpora corpus correlation direct template matching energy Equation error measure estimated gradient fast Fourier transform fftw filter frequency bin pair frequency domain gradient descent harmonic histogram histogram bins HSBSS implementation improvements initialize input signal instrument separation ISMIR-G iteration lateral inhibition learning parameter learning rate learning step locations loudness weights magnitude phase main algorithm matrix method microphones mixing mixture module file monaural musical noise musical piece non-sparseness cost function objective evaluation onset vector optimal output overlap partial derivative peaks problem pseudoinverse ratio reconstruction error reference tracks reverberation RPROP sample segments separation algorithm shift slice solution songs sparse spectral spectrum steering vector Super-SAB template matching tone learning tone search tone waveform update upsampling usually values vector norm window length zero