SPEECH ENHANCEMENT USING DIGITAL

AUDIO EFFECTS AND KALMAN FILTER IN

NOISY ENVIRONMENT

Presented by
U.SANTOSH KUMAR (DECS)
14341D3812
Under the Esteemed Guidance of
Dr. G. Manmadha Rao
Professor, Dept. of ECE
GMRIT
 ABSTRACT

The term Quality of Speech in Speech Enhancement techniques is associated with

Clarity and Intelligibility. Till now due to the variable nature and characteristics of noise with
time and process to process, Speech Enhancement is a difficult problem in Noisy environment.
In this project, proposed a method to improve the quality of speech based on combination of
Digital Audio Effects with Adaptive Kalman Filter when only corrupted speech is available. In
this approach to enhance the Speech content in the Noisy speech signal, Digital audio effects
are used. A Digital Expander generates an audio effect which operates on a low signal level
and create more likely sound characteristics. And further, noise is removed by Auto
Regressive modelled improved adaptive Kalman filter. The performance of the proposed
method with additive color noise is found to be better compared to spectral subtraction, wiener
and Kalman filter methods in terms of input output signal-to-noise ratio and intelligibility in
different real time noisy environment.
Out Lines
Abstract
Objective
Basic Model of Speech Enhancement System
Speech Enhancement Algorithms
Kalman Filter
Digital Audio Effects
Proposed Method
Results, Conclusion and References
 OBJECTIVE

Enhancement of the Speech signal which is degraded by the Non stationary

Additive noise in noisy environment

Improve intelligibility and or quality of speech using proposed method

Performance of Proposed method and Results Analysis

Speech Enhancement System

Speech
Signal
Mixing Fourier
System Windowing
Transform

Noise

Enhanced Inverse
Enhancement
Signal Fourier
Algorithm
Transform
 Speech Enhancement Algorithms

Spectral Subtraction
Wiener Filter Method
Kalman Filter Method
Sinusoidal modelling
Harmonics based techniques
Cascading algorithms
 KALMAN FILTER
Why Kalman Filter?

Easy to implementable on a digital computer

Greater accuracy when compared to Wiener Filter.
It is designed to work with finite data sets.
It makes use of models of the speech and noise productions processes
It can be made to work with non-stationary signals
Provides an estimate of the current parameters using current measurements
and previous parameter estimates.
Speech Enhancement Using Kalman filter
 Design of Kalman Filter
ESTIMATION: State Vector propagation, Parameter Covariance Matrix propagation
UPDATION: Compute Kalman Gain, State Vector update, Parameter Covariance
Matrix update

Prediction(Time Update) Correction (Measurement

Update)
1) Project the STATE 1) Compute the KALMAN
ahead GAIN
k-=Ayk-1+Buk
K=P-kHT(HP-kHT+R)-1
2) Project the Error
Covariance ahead
P-kAPk-1AT+Q 2) Update the estimate
with measurement zk
k=k- + K (zk-H k- )

3) Update the Error Covariance

Pk=(I-KH)P-k
Digital Audio Effects
Classifications:
Basic Digital Audio Filtering Effects:
Equalizer
Time varying Filters: Wah-Wah Filter
Delay Based Effects
Modulation: Amplitude and Phase
Non-Linear Processing: Limiter
Dynamic Range Control

multiplying the delayed input signal x(n)

by a factor g(n)
y(n) = g(n) x(n D)
 Limiter
DRC:
1.Level Measurement2.Static Curve3.Attack and Release Time
 Proposed Method

Considerations:

1.Expanding Factor 0.5

2.AR model Order 20
3.Frame size 25ms
 Results

Clean speech Noisy speech(SNR

5dB)

Spectral Subtraction method Wiener filter

method
 Results

Kalman filter based Speech

Enhancement
Proposed Method Results

Noisy speech Vs Digital Expander output

Proposed Method output and clean speech waveforms

Combination of Digital Audio Effects with Kalman Filter

based Output
Results

Spectrogram of clean & noisy speech signals, Kalman filter method, and proposed method
 Output SNR

Noise Level(in dB)/ Method Spectral Wiener Kalman Proposed

Subtract filter Filter
Clean2 -8.0076 -7.2757 -3.4224 -1.0876
15 -8.2785 -7.2003 -3.3021 -1.4855
Exhibition 10 -8.7255 -7.9856 -3.4796 -1.8703
(dB) 5 -8.8985 -8.8569 -3.5869 -1.6662
0 -9.9985 -9.9723 -4.2569 -1.8839
15 -8.0447 -7.3698 -3.5414 -1.3830
Station (dB) 10 -8.3562 -8.1265 -3.9596 -1.5633
5 -9.7808 -8.9566 -3.9958 -1.2033
0 -9.9025 -9.8991 -4.2650 -2.1383
15 -8.2586 -7.6545 -3.4666 -1.5316
Restraint 10 -8.2826 -7.9856 -3.2491 -1.5001
(dB) 5 -8.7876 -8.5698 -3.5112 -1.4297
0 -9.9957 -9.9848 -3.9676 -1.5051
15 -8.1147 -7.6886 -3.3514 -1.2545
Car 10 -8.7258 -7.9856 -3.4496 -1.4587
(dB) 5 -8.7992 -8.5865 -3.5729 -1.3033
0 -9.9002 -9.8325 -4.2656 -2.2345
 15 -8.2512 -7.0340 -3.7655 -1.4416
Babble 10 -8.5856 -7.8955 -3.8491 -1.4801
(dB) 5 -8.8896 -8.6532 -3.9112 -1.7597
0 -9.8950 -9.7048 -3.9676 -1.7551

15 -8.0006 -7.4527 -3.3568 -1.2405

Airport (dB) 10 -8.2589 -8.1256 -3.3259 -1.0649
5 -8.5866 -8.4569 -3.8026 -1.3569
0 -9.8541 -9.6832 -4.2234 -1.6324

15 -8.0025 -7.1024 -3.7898 -1.3569

Train 10 -8.1256 -7.9856 -3.9887 -1.9456
(dB) 5 -8.6856 -8.3457 -3.9995 -1.9956
0 -9.3254 -9.3251 -4.2587 -2.0001

15 -8.0035 -7.0001 -3.3700 -1.6545

Street 10 -8.2658 -7.4887 -3.3839 -1.4586
(dB) 5 -8.9457 -8.5474 -3.8566 -1.5569
0 -9.6582 -9.5832 -4.2004 -1.6324
 APPLICATIONS
Household Appliances
Speech Recognition Systems
Mobile Phones
Voice over IP
Teleconferencing Systems
Voice Enabled Security Devices
Biomedical Signal Processing,
ATM machines and Computers
 Conclusion
In this Project, Implemented an algorithm based on Adaptive Kalman filter
and Digital Audio Effects, which overcome the disadvantages of earlier
methods.

All these methods are simulated using MATLAB and SNR values of
respective methods are compared. Performance of Proposed method is
analysed in different Noisy Environment like station, train, babble, car, and
restaurant.

Found that the proposed method is giving better SNR values and its
performance is comparatively superior for both stationary and non-
stationary signals.
 REFERENCES
1) Mariyadasu Mathe,Siva Prasad Nandyala, Speech Enhancement using Kalman Filter for white,random and color noise,
IEEE Conf. DOI 10.1109/ICDCSyst.2012.6188703, ICDCS, 2012
2) "Speech Enhancement" by J. Benesty, J. Chen, Eds., and S. Makino, Springer, Berlin, 2005.
3) Sujan Kumar Roy, Wei-Ping Zhu, single channel speech enhancement using subband iterative kalman filter, IEEE
ISCAS, 2016
4) D.V.Anderson and M.A.Clements Audio signal noise reduction using harmonic modeling, in Proc.IEEE Int. conf. Acoust.
ICASSP,1999
5) J.Jensen and J.H.L Hansen,Speech enhancement using a constrained iterative sinusoidal model,IEEE Trans. Speech
Audio Process., vol.9 ,2001.
6) Ramun Schmid; Guido M.Schuster, A Kalman filter based noise suppression algorithm using speech and noise model
derived from spatial information IEEE Signal Processing Conf. 2010.
7) K.K.Paliwal and A.Basu,A speech enhancement method based on Kalman filtering Proceedings of ICASSP87,pp.177-
180. Dallas, TX, USA,1987.
8) B.Chabane and B.DaouedOn the use of Kalman Filter for enhancing speech corrupted by Colored Noise, WSEAS Trans.
On signal processing 2008 Dec.
9) Digital Audio Signal Processing, second edition, by Udo Zolzer
10) Noizeus- A noisy speech corpus for evaluation of speech enhancement algorithms,
http://ecs.utdallas.edu/loizou/speech/noizeus
11) Hu.Y and Loizou.P,Evaluation of objective quality measures for speech enhancement, IEEE Trans. On speech and audio
processing, 2008
THANK YOU