Review on Research Paper & Patent

 Paper Title:SPATIAL AUDIO SEPARATION USING PROJECTIONS

 Authors: Derry FitzGerald, Antoine Liutkus, Roland Badeau

A projection-based method for the unmixing of multichannel audio signals into their different constituent spatial objects. Here, spatial objects are modelled using a unified framework which handles both point sources and diffuse sources. These linear combinations consist of observations where some spatial objects are cancelled or enhanced. The algorithm was also informally tested on a number of commercial recordings.


Link:http://ieeexplore.ieee.org/document/7471632/authors



Patent Title:  Audio source separation

Authors: WANG, David S
 
The method includes determining a spatial parameter of an audio source based on a linear combination characteristic of the audio source and an orthogonality characteristic of two or more audio sources to be separated in the audio content. The method also includes separating the audio source from the audio content based on the spatial parameter. Corresponding system and computer program product are also disclosed. It will be appreciated that the example embodiments disclosed herein are not to be limited to the specific embodiments disclosed and that modifications and other embodiments are intended to be included within the scope of the appended claims. Although specific terms are used herein, they are used in a generic and descriptive sense only and not for purposes of limitation.

DSP processor : Hardware implementation

We were given the TMS320F28335 DSP kit with JTAG. We were taught to use the Code Composer Studio v4,0 to write the program on to the DSP processor. The Code Composer Studio allows real time debugging of the program in running mode. The various features of DSP processor can be exploited using this feature. We performed convolution, correlation, DFT, FFT, OAM and OSM on the processor kit. Thus, we got a hands-on experience of the DSP kit.

FIR Filter Design Using FSM

Experiment No.8

In this experiment, we were required to design a digital FIR filter using Frequency Sampling Method (FSM). Scilab was used to implement it.

We implemented a digital FIR filter using Frequency sampling method. The input specifications were taken as usual and the magnitude response was plotted. 

In frequency response of the obtained output signals of Frequency Sampling Method, ripples in the Stop Band obtained are of decreasing amplitudes. Phase plot of HPF and LPF are same if the order of both of them are same. Phase varies linearly with frequency , hence output will not be distracted.  Also that as order increases number of side lobe increases.

FIR Filter Design Using Windows Function

Experiment No.7

The user was prompted to input values like Attenuation in Stop band (As) and Pass band (Ap) as well as Pass band frequency, Stop band frequency and sampling frequency.A low pass and Band pass filter was designed.

The program calculated which windowing function would be most suitable for the given input specifications.Thus, we performed the experiment for two window functions: Bartlett and Hanning.

 The magnitude and phase plot of both the filters was plotted using scilab. In this method, the desired impulse response is multiplied with window function w(n) to obtain h(n) which after Z-transfrom  gave H(z). The phase  plot being linear, there will be no distortion at the output.

We learnt that Hamming Window function gives more attenuation in stop band than Hanning window, hence it is a better window function.

CHEBYSHEV FILTER DESIGN

Experiment No.6

we observe that the filter has equiripple behavior in passband and monotonic behavior in stopband 

Order of the butterworth filter is lower than butterworth and transition band is narrow than butterworth filter 

We searched for inbuilt code to plot magnitude response. We tried different function and finally we got its function to plot magnitude response as"cheb1mag".

We performed both low pass filter and high pass filter design of chebyshev filter.

DIGITAL BUTTERWORTH FILTER DESIGN

Experiment No. 5

The butterworth filter design was done for Low Pass and High Pass. It was done with specific input parameters like As, Ap , Pass Band frequency, Stop Band Frequency and Sampling frequency. H(z) for each were calculated and so was the order of the filter. The theoretical and observed values were compared. Maginitude spectrum and pole zero plot was drawn and it was observed that butterworth filters are monotonic in its pass band and stop band.

It was noticed that the magnitude response smooth in both pass band and stop band. With higher order, the magnitude response became sharoer and started resembling the ideal filter more closely.

Overlap Add & Save Method

Experiment 4

When the data size and rate is high, we filter the long data sequence using Overlap Add Method and Overlap Save Method.In OAM, convolution of smaller groups of input with the second input is obtained. So we get convolution for each group which gets overlap depending upon the length of the input signal. In OSM, we divide the output instead of input signals and use only limited input required to calculate that output. Hence the memory requirement is reduced. Both the methods require the same amount of computations and hence the same memory.

Fast Fourier transform

Experiment 3

For finding DFT for large signal we use FFT which divides N point signal into two equal parts with each part consists of alternate values.
On comparing the arithmetic computing properties for complex addition, Complex multiplication, real addition and real multiplication of FFT with DFT, we observed that number of calculations were less for FFT, thus making FFT much faster than DFT. The reason for fast computation in FFT is parallel processing.

Discrete Fourier transform

Experiment 2

Discrete Fourier Transform (DFT) is frequency sampling of Discrete Time Fourier Transform (DTFT). 
DFT of 4 point sequence was calculated and then. we padded the 4-point signal with 4 zeroes and observed the change in magnitude spectrum from the 4-point signal. DFT results are periodic with period equal to length of the input signal.
If the input signal is expanded, its Magnitude Spectrum is compressed. The total number of multiplications and additions in DFT is higher and hence DFT is computationally slow.

Convolution and Correlation Algorithms

Experiment 1

In this experiment  Convolution (Linear Convolution, Circular Convolution and Linear Convolution using Circular Convolution) and Correlation (Auto Correlation and Cross Correlation) has been performed.
Convolution is a mathematical way of combining two signals to form a third signal. It is the single most important technique in Digital Signal Processing. Using the strategy of impulse decomposition, systems are described by a signal called the impulse response. Convolution is important because it relates the three signals of interest: the input signal, the output signal, and the impulse response. 

Linear Convolution is used for non-periodic signals and Circular Convolution is used for periodic signals. The output of Circular Convolution is aliased. Application of Convolution is to find output of the system. We observed that for Linear Convolution the length of output signal is N=L+M-1 and For the circular that length is N= MAX(l,M).

The output of Auto Correlation is the value of the output at n=0 gives the energy of the signal. Auto Correlation of delayed signal is same as that of original signal. The output was both-sided.