Room Transfer Function Estimation and Room Equalization in Noise Environments

Deokgyu Yun.et al. Int. Journal of Engineering Research and Applications www.ijera.com
ISSN: 2248-9622, Vol. 6, Issue 4, (Part - 4) April 2016, pp.38-40
www.ijera.com 38|P a g e
Room Transfer Function Estimation and Room Equalization in
Noise Environments
Deokgyu Yun*, Min-Jeong Kim**, Taeuk Kim**, Siyeon Nam**,
Se Rin Hong**, Seung Ho Choi**
* (Department Of Electronic Engineering, Seoul National University Of Science And Technology, Korea)
** (Department Of Electronic And IT Media Engineering, Seoul National University Of Science And
Technology, Korea,)
Corresponding author: Seung Ho Choi (shchoi@seoultech.ac.kr)
ABSTRACT
Audio quality in listening situation is degraded by indoor room reverberation. Room equalization can be used to
increase the audio quality by applying the inverse transfer function to the input audio signals. In noise
environments, however, it is hard to exactly measure the room transfer function. In this work, we developed the
techniques to measure the room transfer function in indoor noise environments and to enhance the audio quality
by room equalization. From the experimental results, we showed that the proposed techniques can be
successfully used in indoor noise environments.
Keywords – room transfer function, room equalization, reverberation, noise environments.
I. INTRODUCTION
Audio quality in listening situation is
degraded by indoor room reverberation and can be
increased by using room equalization that applies the
inverse of room transfer function (RTF) to input
signals. First, we used the short-time spectral
masking method for RTF estimation in noise
environments [1]. Next, room equalization
techniques are developed for the audio quality
enhancement. The audible frequency range is
separated into unequal segments by one-third octave
band filters. In addition, the estimated RTF is
smoothed in frequency domain, and the gain
parameters for the equalization are obtained by
inversing the average spectral magnitude at each
band and are normalized.
First, Section 2 introduces the RTF
estimation method in noisy environments. Section 3
describes room equalization based on smoothed RTF.
Section 4 shows the experimental results. The last
Section concludes this work.
II. ROOM TRANSFER FUNCTION
ESTIMATION IN NOISE
ENVIRONMENTS
For the room equalization, first, RTF
should be measured. In this work, we used sine
sweep signal for the accurate RTF measurement [1].
As shown in Fig.1, the sine sweep signal )(nx is
convolved with )(nh that is room impulse response.
The not processed signal )(ny np is a noisy signal .
Figure 1. RTF measurement method [1]
Fig. 1 shows the process for the RTF
measurement. The method divides )(ny np into 50%
overlap frames and applies the Hanning window.
Then short-time Fourier transform (STFT) is applied
to obtain )(kY np . Next, )(kY np is masked to obtain
the frequency response of the sine sweep signal. The
masker has a different range and location for each
frame. )(max iK is the frequency bin index of largest
frequency component in )(kX when the frame
index is i . The masker ),( kiM is described as in
Eq. (1).


 

otherwise
ikiKi
kiM
,0
)()()(,1
),(
2max1
(1)
The masked spectral components are expressed in
Eq. (2).
),(),(),( kiMkiYkiY npprop  (2)
Likewise, ),( kiY prop is divided into 50% overlap
frames and the Hanning window is applied.
Afterward, inverse STFT is performed to obtain the
noise-suppressed signal )(ny prop .
RESEARCH ARTICLE OPEN ACCESS

III. ROOM EQUALIZATION
In this section, we describe the room
equalization techniques based on RTF estimated by
the short-time spectral masking method. When the
RTFs are known a priori or are capable of being
accurately estimated, this approach has been shown
to achieve high inverse filtering performance [2, 3].
However, in actual acoustic environments, there are
disturbances that affect the inverse filtering
performance. In some audio applications (e.g., sound
reproduction systems in train stations or other large
spaces) where audio quality is an issue, an
equalization filter is commonly used to compensate
for the frequency response of the room. The
equalization performance then depends on the model
from which the inverse filter is derived. In our work,
the equalization filter is one-third octave band filter
which is composed of 27-octave bands that first
frequency doubled every 3rd octave at a sampling
frequency 48000Hz. Also, the center frequency of
first band is about 47Hz. Filter gains are obtained by
averaging the smoothed RTF. In this work, we use
the complex smoothing method [4]. The complex
smoothing operation may be described as a circular
convolution as in Eq. (3).
)()mod)(()()()(
1
0
iWNikHkWkHKH sm
N
i
smCS 


 (3)
where the symbol  denotes the operation of
circular convolution and )(kH is transfer function,
)(kW sm is a spectral smoothing function having the
general form of a low-pass filter. )( KH CS is the
result of complex smoothing of )(kH . By using the
filters that have gains based on the smoothed inverse
RTF, the overall RTF is equalized. audio quality, in
such scenarios, is enhanced by performing the room
equalization.
IV. EXPERIMENT AND RESULTS
From the experiments, the estimated RTF
and inverse of estimated RTF are shown in the Fig. 2.
(a) The estimated RTF function
(b) The inverse of estimated RTF function
Figure 2. The estimated RTF and the inverse of the
RTF
The processed signal by the proposed room
equalization becomes similar to original signal. In
this experiment, we used log spectral distance (LSD)
for objective assessment. The LSD is the difference
between reference signals and signals obtained in the
frequency region, and can be calculated as
  

2
2
2
))
),(
),(
log(10(
1
1
(
1
1
kiY
kiY
k
N
Ni
M
M
LSD
ref
(4)
where M is the number of frames. As shown in the
Table 1, the developed room equalization techniques
improved audio quality. The table shows the LSD
comparison results between original signal and not
processed or processed signals, respectively. Two
male and two female speech signals were used for
the experiments. The sampling frequency was 48
kHz and the length of signals was 4 seconds.
Table 1. LSD (dB) between original signal and not
processed or processed signals
signal not processed processed
Male 1 29.86 25.84
Male 2 29.73 25.78
Female 1 29.12 25.08
Female 2 28.87 25.13
V. CONCLUSION
In listening situations, audio quality can be
degraded by reverberation. First, we developed the
techniques to measure the room transfer function in
noise environments. Then, room equalization is used
to enhance the audio quality by applying the inverse
transfer function to the input audio signals. From the
experiments, it is shown that the room equalization
techniques can improve the audio quality in
reverberant indoor room environments.

ACKNOWLEDGEMENTS
This study was supported by the Research
Program funded by the Seoul National University of
Science and Technology.
REFERENCES
[1]. Deokgyu Yun, Jonghan Joo, Seunghyon
Na, Min-jeong Kim, Yeoung Hwangbo and
Seung Ho Choi, “A Short-Time Spectral
Masking Method for Room Tranfer
Function Measurement Based on Sine
Sweep Signal in Noisy Environment,” The
2nd
International Joint Conference on
Convergence, pp. 220-221, Jan. 2016.
[2]. P. A. Nelson, F. Orduna-Bustamante, and H.
Hamada, “Inverse filter design and
equalization zones in multichannel sound
reproduction,” IEEE Trans. Speech Audio
Process., vol. 3, no. 3, pp. 185 – 192, May
1995.
[3]. Tobias Corbach, Adrian von dem
Knesebeck, Kristjan Dempwolf,
“Automated Eeualization for Room
Resonance Suppression,” Proc. Of the Int.
Conf. on Digital Audio Effects, Sep. 2009.
[4]. Panagiotis Hatziantoniou, John
Mourjopoulos, “Generalized Fractional-
Octave Smoothing of Audio and Acoustic
Responses,” Journal of the Audio
Engineering Society, Vol. 48, No. 4, pp.
259-280, Apr. 2000.

Room Transfer Function Estimation and Room Equalization in Noise Environments

More Related Content

What's hot (19)

Similar to Room Transfer Function Estimation and Room Equalization in Noise Environments (20)

Recently uploaded (20)

Room Transfer Function Estimation and Room Equalization in Noise Environments