Unlocking the Potential of the Cloud for IBM Power Systems
Speech measurement using laser doppler vibrometer
1. Guided By: Presented By:
Asif Ali Shamil. C
Lecturer in E.I Roll no: 68
E.I
2. Introduction
Speech measurement with LDV
Principe of LDV
Measurement Setup
Problem formulation
Speech Enhancement Algorithm
Speckle noise suppression
LDV-Based time frequency VAD
Spectral gain modification
Experimental Results
Conclusion
3. Achieving high speech intelligibility in noisy
environments is one of the most challenging and
important problems for existing speech-
enhancement and speech-recognition systems.
Recently, several approaches have been
proposed that make use of auxiliary non acoustic
sensors, such as bone and throat- microphones.
Major drawback of most existing sensors is the
requirement for a physical contact between the
sensor and the speaker.
Here present an alternative approach that
enables a remote measurement of speech, using
an auxiliary laser Doppler vibrometer (LDV)
sensor.
4.
5. fd(t) = 2ν(t) cos(α)/λ
ν(t)=> instantaneous throat-vibrational
velocity
α => Angle between the object beam and the
velocity vector
λ =>laser wavelength.
LDV-output signal after an FM-demodulator
is
Z(t) = fb + [2Av cos(α)/λ].cos(2πfvt). (1)
9. let y(n) =x(n) + d(n)
y(n)-observed signal in the acoustic sensor.
x(n) -Speech signal.
d(n)-Un correlated additive noise signal.
In the STFT domain,
Ylk = Xlk + Dlk
Where l= 0, 1, . . . is the frame index.
k = 0, 1, . . . , N − 1is the frequency-
bin index.
10. Use overlapping frames of N samples with a
framing-step of M samples.
Let H0lk and H1lk indicate, respectively, speech
absence and presence hypotheses in the
time-frequency bin (l, k), i.e.,
H0lk: Ylk = Dlk
H1lk: Ylk = Xlk + Dlk.
X̂lk = GlkYlk.
11. The OM-LSA estimator minimizes the log spectral
amplitude under signal presence uncertainty
resulting in,
Glk = {GH1lk}ˆPlk.Gminˆ1−Plk .
Where,
GH1lk is a conditional gain function given H1lk &
Gmin<< 1 is a constant attenuation factor.
Plk is the conditional speech presence
probability.
12. Denoting by ξlk and γlk we
get,
-Priori SNR
-Posteriori SNR
is the a priori probability for speech
absence,
13. Speckle-Noise Suppression
The output of the speckle-noise detector is,
Wl(n) = Gl Zl(n)
Where Gl= Gsmin<<1 for Il = 1(speckle noise is
present)
Gl = 1 otherwise.
14.
15. -Represents the noise-estimate bias
-Smoothed-version of the power spectrum
Then, we propose the following soft-
decision VAD:
16.
17. Speech in a given frame is defined by
We attenuate high-energy transient components to the level
of the stationary background noise by updating the gain
floor to
-Stationary noise-spectrum estimate
-Smoothed noisy spectrum
18.
19. Speckle noise was successfully attenuated
from the LDV-measured signal using a
kurtosis-based decision rule.
A soft-decision VAD was derived in the
time-frequency domain and the gain function
of the OM-LSA algorithm was appropriately
modified.
The effectiveness of the proposed approach
in suppressing highly non-stationary noise
components was demonstrated.
20. I. Cohen and B. Berdugo, “Speech enhancement for nonstationary noise
environment,” Signal Process., vol. 81
T. F. Quatieri, K. Brady, D. Messing, J. P. Campbell, W. M. Campbell, M. S.
Brandstein, C. J.Weinstein, J. D. Tardelli, and P. D. Gatewood, “Exploiting
nonacoustic sensors for speech encoding,”
T. Dekens, W. Verhelst, F. Capman, and F. Beaugendre, “Improved speech
recognition in noisy environments by using a throat microphone for accurate
voicing detection,” in 18th European Signal Processing Conf. (EUSIPCO),
Aallborg, Denmark, Aug. 2010, pp. 23–27
M. Johansmann, G. Siegmund, and M. Pineda, “Targeting the limits of laser
doppler vibrometry,”
http://www.metrolaserinc.com
