Paper Review : Effect of Low Pass Filtering of Digital Frequencies in Quantized Signals

 Effect of Low Pass Filtering of Digital Frequencies in Quantized Signals

A linear array of electrodes was used to represent the discrete digital frequencies in the processed signal. A specific electrode in the array would be activated according to the number of quantizing intervals in a given high portion of the quantized waveform. This corresponds to performing a frequency-to-distance mapping of speech similar to what is done by the basilar membrane of the cochlea. If certain frequencies are not displayed, then this reduces the number of electrodes needed and corresponds to a filtering operation. The effect of reducing the number of electrodes needed by high-pass filtering the processed signal has already been studied. The current research addresses the
trying to reduce the number of electrodes low-pass filtering the processed speech.This was
measured using the normal aural channel.


Paper Title: The Effect of Low-pass Filtering of the Digital Frequencies in Quantized Speech.
Authors: Edward M. O'Brien and David J. Pogue
Published at: IEEE ENGINEERING IN MEDICINE & BIOLOGY SOCIETY 10TH ANNUAL INTERNATIONAL CONFERENCE--I107
Publisher: IEEE

Patent Review : US5212764

Patent Review

The objective of the present invention is to provide a noise eliminating apparatus capable of effectively eliminating background noise components from a speech signal containing the same irrespective of the levels of the background noise components.Another objective of the present invention is to provide a noise eliminating apparatus in which coefficients for eliminating noise are renewed as many times aspossible so that background noise contained in a speech signal is effectively eliminated in response to a change of noise.The above-mentioned objectives of the present invention are achieved by a noise eliminating apparatus comprising: 

First feature extracting means for individually extracting a feature quantity X(f) (f = l, 2, . . . , n) of an input speech containing a noise for n channels 

(n is an integer);

Second feature extracting means for individually extracting a feature quantity N(f) of the noise for the 11 channels; and noise eliminating means, 

coupled to the first and second feature extracting means, for deriving a first noise eliminating variable k(f) and a second noise eliminating variable d(f) from the feature quantity X(f) of the input speech containing the noise provided for each of the n channels and the feature quantity N(f) of the noise provided for each of the n channels and for generating a feature quantity S(f) of a noise-eliminated input speech.

DSPP EXPERIMENT NO. 09

Basic Operations using DSP Processor

DSP Processors are capable of handling large mathematical operations easily while using real-time signals for manipulation of data.In this practical, DSP C2000 processor is used. The coding platform used was Code Composer Studio.Some basic operations like addition, subtraction, logical shifting are performed. We also implemented Fast Fourier Transform(FFT) algorithms.The difference in implementing on hardware is that we have to reference the registers too while writing the code.If implementation is done using C language, we do not require to reference the registers.

DSPP EXPERIMENT NO. 08

Digital FIR Filter Design using Frequency Sampling Method

Frequency Sampling Method (FSM) is another way of designing linear phase FIR filter.We have used SCILAB software for practical implementation. The formula of DFT and IDFT were put in code.Also filter parameters were taken as input from the user.The plot function enables the user to verify the accurateness of the designed filter.

DSPP EXPERIMENT NO. 07

Digital FIR Filter Design using Windowing Method

A linear phase finite impulse response(FIR) filter was designed using Scilab. Filters like stop band and pass band attenuation and stop band and pass band frequency are taken as input from the user.There are different window functions that can be used like Hamming Window, Bartlet Window, Hanning Window, Blackman Window,etc. The window function used in this experiment is Hanning window.the formula is input in the Scilab code and then the program is executed 

DSPP EXPERIMENT NO. 06

Digital Chebyshev filter Design

Chebyshev filter has ripples in passband and is monotonic in in stop band.The number of peak and valleys in the characteristic curve determines the order of the filter.Same procedure used for design of digital butterworth filter is used for design of digital chebyshev filter. Poles of chebyshev filter lie on ellipse while poles of butterworth filter lie on the unit circle.

DSPP EXPERIMENT NO. 05

Digital Butterworth Filter Design

In this experiment we used Scilab software to simulate behaviour of digital butterworth filters.It provides an easy interface for coding and can also generate frequency response curves.A butterworth filter has a maximally flat magnitude frequeny response curve . Bilinear transformation method was used to design the filter. Filter parameters like passband attenuation,stopband attenuation,passband frequency and stopband frequency were given as input by the user.The order and transfer function of filter was displayed and magnitude response of filter was displayed on the screen. 

DSPP Experiment No. 04

               Overlap Add Method and Overlap Save Method

We use overlap add method(OAM) and overlap save method(OSM) when input signal x(n) is given to digital FIR filter.OAM and OSM are linear filtering methods.In this practical,a 13 point input sequence is taken and output is found out using C program using OAM and OSM for same input sequence.The outputs obtained using both the methods are same.

In OAM method,input sequence is decomposed and then linear convolution is performed on these signals individually.The outputs obtained are then combined together and displayed.In OSM the signals are decomposed and then circular convolution is performed.The input is decomposed in OSM in such a way that there are some value from previous sequence.The outputs obtained are then combined inbto a single sequence by removing the repeated number of values.In this practical,x(n) was a 13 point sequence and h(n) was 4 point sequence.

DSPP Experiment No. 03

                            FAST FOURIER TRANSFORM 

Fast fourier transform(FFT) is a fast method of computation than Discrete fourier transform(DFT) as FFT has less number of additions(both real and complex) and multiplications(both real and complex) involved.Operational efficiency of FFT is high as it uses parallel processing.Also the output of FFT is in bit reversed form.In this practical,two cases were considered by taking 4 point and 8 point input sequences.The output was found using C program which implemented DITFFT(Decimation in time FFT) algorithm.

Experiment No. 02

                  DISCRETE FOURIER TRANSFORM

Discrete fourier transform(DFT) gives periodic results.It is frequency sampled version of discrete time fourier transform.DFT is a slow algorithm.

In this experiment,initially a 4 point sequence was taken and it's DFT was found out using C program.Then this same sequence was appended by 4 zeros to form a 8 point sequence.It's DFT output was found out.Then magnitude spectrum was plotted for both output sequence.It was observed that the magnitude spectrum of 8 point input sequence was more accurate than 4 point input sequence.It was because the resolution of magnitude spectrum of 8 point sequence was higher.

DSPP Experiment No. 01

                              Convolution and Correlation

Linear convolution is used to find output of the system whereas correlation is used to find degree of similarity of two signals.These concepts are given to us in theory of digital signal processing.In this experiment,for linear convolution,x(n) is taken as input along with impulse response h(n).The output is y(n).For length of output signal y(n) of convolution operation,we use the formula N>=L+M-1 .For length of output signal of correlation operation we use the formula N=max(L,M).In this practical,the codes for these operations were studied and then implemented.Linear convolution,circular convolution,linear convolution using circular convolution,auto-correlation and cross correlation were performed.The output obtained was then verified with the answers obtained by theoritical calculations.