Digital Signal Processing: PBL Approach: Theme: Speech Processing Title

Download as pptx, pdf, or txt
Download as pptx, pdf, or txt
You are on page 1of 15

Digital Signal Processing: PBL approach

Theme: SPEECH PROCESSING


Title: VOICE ENHANCEMENT USING SPECTRAL SUBTRACTION

Course: Digital Signal Processing


Course Code: 23EECC303 Team Details
Sl.No. Roll No. Div SRN Name
Semester: V 1 403 D 01FE21BEC175 G.Gyaneshwar Rao

Credits: 4 [2-0-2] 2 404 D 01FE21BEC176 Sanjana Madiwalar

Hours/Week: 6 3 406 D 01FE21BEC178 Yaseer Mulla

01FE21BEC222
Faculty Mentor: 4 442 D Rakshita P.S

08/11/2024 School of ECE 1


Title:

Content

 Problem Statement
 Introduction
 Literature Survey
 Functional Block Diagram
 Data Sets/Data Acquisition
 Methodology
References

08/11/2024 School of ECE 2


Title:

Problem Statement

VOICE ENHANCEMENT USING SPECTRAL SUBTRACTION

08/11/2024 School of ECE 2


Title:

Introduction
Voice enhancement using spectral subtraction
is a technique aimed at improving audio signal
quality, particularly in noisy environments. By
analyzing the frequency spectrum of the audio
input, the method identifies and isolates
components related to background noise.
These noise components are then subtracted,
reducing unwanted interference and
enhancing the clarity of the desired speech
signal.

08/11/2024 School of ECE 2


Sl. Title Methodology/ Merits Demerits Gaps
No Algorithm
1 Noise reduction algorithms, 1. Human Auditory 1. Requirement of
Accurate Pitch
1. Lack of solutions for the
Digital Signal Processing frequency dependent System as Prototype
2. Estimation of
poor performance of
Based Speech amplification, amplitude 2. Algorithmic hearing aids in noisy
compression, voice activity Enhancements Limitations of Adaptive environments,
Enhancement Techniques detection algorithms, and 3. Tailored Auditory Noise Cancellation 2. The need for accurate
for Hearing Impaired Technique
frequency shaper using a digital Experience 3. Assumption of pitch estimation
People filter bank approach or FFT based 4. Future Trends in Stationary Noise in 3. The practicality of
approach. Hearing Aids Spectral Subtraction adaptive noise cancellation
Anticipated Technique techniques
Innovations 4. Possible Non- 4. The potential for
Stationarity of Noise in distortions due to
Delay-Based Single amplitude compression.
Channel Adaptive Noise
Cancellation
2 Speech Enhancement Multi-band Spectral Subtraction 1. Introduction of 1. Algorithm's 1. The application of
based on Fractional Method, Enhancement and Fractional Fourier Dependency on Prior Fractional Fourier
Bandwidth Compression of Noisy Transform (FRFT) Speech Spectrum Transform (FRFT) for
Fourier transform Speech, Filtering of Colored Noise, 2. Superior Denoising 2. Estimates Increased speech enhancement
Subband Kalman Filtering, Iterative Effect Computational and denoising is
filter-based speech and sequential 3. Robustness Under Complexity discussed
Kalman enhancement algorithms, Various Noisy 3. Degradation in Low 2. Its effectiveness over
Signal subspace approach, and Environments and Signal-to-Noise Ratio traditional methods, and
Fractional Fourier Transform (FRFT) SNR Conditions or Non-Stationary its implementation and
filtering. 4. Low Computational Noise Conditions experimental results.
Complexity Significant 3. It also compares FRFT
Improvement in with other denoising
Signal-to-Noise Ratio techniques.
(SNR)

08/11/2024 School of ECE 5


3 A Review of Speech Signal spectral subtraction 1. Improved Speech 1. Loss of No technical gaps

Title:
Enhancement Technique algorithms, non-linear
spectral subtraction,
Quality.
2. Enhanced Intelligibility 2.
Intelligibility
Computational
adaptive noise 3. Noise Suppression Complexity

Literature Survey
cancellation, multisensor 4. Adaptive Noise 3. Dependence on
beamforming, Wiener Cancellation Noise Estimation
filtering, Kalman filtering,
linear predictive coding.

4 Speech Enhancement Magnitude spectral 1. The proposed hybrid 1. The proposed 1. Does not provide a
subtraction, 2-D Wiener filter outperforms method may not detailed analysis of
Using 2-D Fourier filtering and a hybrid filter other methods in work well for the trade-offs
Transform that combines 1-D and 2-D objective tests. other types of between speech
Wiener filters 2. The hybrid filter noise or speech quality and
particularly effective in signals computational
reducing white noise 2. The objective test complexity
distortion in speech used may not fully 2. Does not discuss
signals. capture subjective the potential
3. Practical solutions for perceptions of impact of the
improving speech speech quality. proposed method
quality in noisy 3. Does not give on speech
environments. detailed analysis of recognition
the computational
complexity or real-
time performance
08/11/2024 School of ECE of the proposed 2
5 Signal Enhancement by 1.Signal Enhancement by 1. Clean Signal No such demerits No technical gaps or limitations
Title:
Time-Frequency Peak
Filtering'
Time-Frequency Peak
Filtering
Recovery at Low
Signal-to-Noise
2. The method is Ratios (SNR)

Literature Survey
implemented using the 2. Effective Testing
pseudo Wigner- on Simulated and
Ville distribution. Real Data
3. Significant
Enhancement of
Signals

6 Improve Speech Application of Weiner 1. Exploration of 1. Lack of Detailed 1. Need for More In-Depth
filtering for speech Speech Discussion on Specific Analysis of Limitations and
Enhancement Using enhancement Enhancement with Challenges or Challenges in Applying
Weiner Filtering Wiener Filtering Drawbacks of Wiener Wiener Filtering for Speech
2. Addressing Filtering for Speech Enhancement
Challenges and Enhancement - 2. Lack of Comprehensive
Factors Influencing 2. More Comprehensive Comparison of Different
Speech Quality Comparison of Tested Estimators of the A Priori
Estimators of the A Signal-to-Noise Ratio
Priori Signal-to-Noise
Ratio

7 Configurable Digital DFT-based transforms 1.Novelty 1.Limited scope 1.Lack of detailed methodology
Hearing Aid System with domain methods such as 2.More flexibility 2.Lack of comparison 2. Limited generalizability
spectral subtraction 3.Enhances speech 3.Lack of real-world 3. Limited evaluation metrics
Reduction of Noise for quality testing
Speech Enhancement
Using Spectral
Subtraction Method and
Frequency Dependent
Amplification.

08/11/2024 School of ECE 2


8 Enhancement of Speech Implementation of noise - Faster Sound No such demerits No technical gaps or limitations
Signals for Hearing Aid reduction systems, Processing
including fixed filters, - Noise Reduction
Devices using Digital adaptive filters, and Capabilities
Signal Processing spectral analysis - Potential for
techniques Individualized
Programming

9 Improve Speech Application of Weiner 1. Exploration of 1. Lack of Detailed 1. Need for More In-Depth
filtering for speech Speech Discussion on Analysis of Limitations and
Enhancement Using enhancement Enhancement with Specific Challenges Challenges in Applying
Weiner Filtering Wiener Filtering or Drawbacks of Wiener Filtering for Speech
2. Addressing Wiener Filtering for Enhancement
Challenges and Speech 2. Lack of Comprehensive
Factors Influencing Enhancement - Comparison of Different
Speech Quality 2. More Estimators of the A Priori
Comprehensive Signal-to-Noise Ratio
Comparison of
Tested Estimators of
the A Priori Signal-
to-Noise Ratio

08/11/2024 School of ECE 8


Title:
Data Sets/Data Acquisition
LibriMix is an open source dataset for source separation in noisy
environments. It is derived from LibriSpeech signals (clean subset) and
WHAM noise. It offers a free alternative to the WHAM dataset and
complements it. It will also enable cross-dataset experiments.

The number of sources in the mixtures.


The sample rate of the dataset from 16 KHz to any frequency below.
The mode of mixtures : min (the mixture ends when the shortest source
ends) or max (the mixtures ends with the longest source)
The type of mixture : mix_clean (utterances only) mix_both (utterances +
noise) mix_single (1 utterance + noise)

08/11/2024 School of ECE 2


Title:
Functional Block Diagram

gggg
g
08/11/2024 School of ECE 2
Title:
Methodology

 SPECTRAL SUBTRACTION is a noise reduction technique for audio


signals.
 It involves transforming the signal into the frequency domain using
the Discrete Fourier Transform (DFT), estimating the noise spectrum,
and subtracting it from the original spectrum.
 The result is an enhanced signal with reduced background noise.
 The effectiveness depends on careful tuning of parameters, such as
the subtraction factor (alpha).

08/11/2024 School of ECE 2


Title:
WEINER FILTER
Sampling WEINER
(here it depends
on input) FILTER

Input Signal Moving Output Signal


Average ZOOM IN
Filter
STEPS Weiner filter
1 Noise power
ZOOM IN
STEPS Moving Average Filter
2 Audio power
3 Wiener filter =
1 Window size 1 - (noise power / audio power)
2 Apply a moving average 4 Apply the Wiener filter element-wise
filter with the defined to the noisy signal: filtered signal = y
window size to the filtered * wiener filter
signal

08/11/2024 School of ECE 2


Title:
Spectral Subtraction
Sampling
and WEINER
Input Signal window FILTER Output Signal
size
Input audio signal output audio signal

Estimate Combine
Spectral
FFT Noise Magnitude Inverse FFT
Subtraction
Spectrum and Phase

ZOOM IN

Applies spectral
subtraction to the
magnitude spectrum
of the signal.

08/11/2024 School of ECE 2


Title:
References
1. Ing Yann Soon and Soo Ngee Koh, “Speech enhancement using 2-D Fourier transform,” in
IEEE Transactions on Speech and Audio Processing, vol. 11, no. 6, pp. 717-724, Nov. 2003,
doi: 10.1109/TSA.2003.816063.
2. B. Boashash and M. Mesbah, “Signal enhancement by time-frequency peak filtering,” in IEEE
Transactions on Signal Processing, vol. 52, no. 4, pp. 929-937, April 2004, doi:
10.1109/TSP.2004.823510.
3. R. Vullings, B. De Vries and J. W. M. Bergmans, “An Adaptive Kalman Filter for ECG Signal
Enhancement,” in IEEE Transactions on Biomedical Engineering, vol. 58, no. 4, pp. 1094-
1103, April 2011, doi: 10.1109/TBME.2010.2099229.
4. B. Saha, S. Khan, C. Shahnaz, S. A. Fattah, M. T. Islam and A. I. Khan, “Configurable Digital
Hearing Aid System with Reduction of Noise for Speech Enhancement Using Spectral
Subtraction Method and Frequency Dependent Amplification,” TENCON 2018 – 2018 IEEE
Region 10 Conference, Jeju, Korea (South), 2018, pp. 0735-0740, doi:
10.1109/TENCON.2018.8650450.

08/11/2024 School of ECE 2


Thank You

08/11/2024 School of ECE 2

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy