This paper proposes a noise-biased compensation of minimum statistics (MS) method using a nonlinear function and
Noise estimation algorithms are essential components of many modern mobile communication, speech recognition, and human computer interaction systems for speech enhancement [
Current single microphone speech enhancement methods belong to two groups, namely, time domain methods such as the subspace method and frequency domain methods such as the spectral subtraction (SS) [
Several recent studies have proposed noise estimation schemes for unknown noise signals [
To solve this problem, we propose a combined adaptive factor based on a sigmoid function and
We evaluate the performance of the proposed algorithm for nonstationary noise and various noise environments. The improvement can be confirmed in the segmental SNR and the Itakura-Saito Distortion Measure (ISDM) [
The noisy speech signal
The MS algorithm relies on the fact that the noisy power spectrum often becomes equal to the noise power spectrum during periods of speech pauses [
The estimated noise power based the MS algorithm [
The two-state model of speech events can be represented as a binary hypothesis model [ if else
where
In this section, we propose a method that combines the adaptive factor based on the sigmoid function and the
First, we can detect the adaptive factor by requiring the smoothed power spectrum
(a) Plot of the adaptive factor
In this subsection, our method uses another adaptive parameter to control the trade-off between speech distortion and residual noise for suppressing the estimated noise in a highly nonstationary and varying noisy environment. The autocontrol parameter is controlled by
The estimated clean speech power spectrum can be represented as shown in (
The noisy signals used in our evaluation were taken from the NOIZEUS database [
Comparison between the noisy signal, noise estimated by MS, and noise estimated by the proposed method in restaurant 5 dB noisy environment.
Figure
Comparison between the noisy signal, noise estimated by minimum statistics (MS), and noise estimated by the proposed method in highly nonstationary noisy environments.
The spectrum of the clean signal is given in Figure
Frequency domain results of speech enhancement for exhibition noise 5 dB SNRs in noisy environments. (a) Original spectrogram, (b) spectrogram using MS with SS method, (c) spectrogram using the MCRA with SS method, and (d) spectrogram using the proposed method.
Tables
Objective evaluation and comparison of the proposed method segmental SNR values.
Noise | Method | SNR | |||
---|---|---|---|---|---|
0 (dB) | 5 (dB) | 10 (dB) | 15 (dB) | ||
White | MS | 4.27 | 8.77 | 12.83 | 16.57 |
MCRA | 5.08 | 9.99 | 13.56 | 17.15 | |
Proposed |
|
|
|
|
|
|
|||||
Car | MS | 3.44 | 7.48 | 12.01 | 16.10 |
MCRA | 4.92 | 7.93 | 11.85 | 16.42 | |
Proposed |
|
|
|
|
|
|
|||||
Babble | MS | 1.83 | 6.00 | 11.17 | 15.03 |
MCRA | 3.73 | 6.79 | 10.52 | 16.22 | |
Proposed |
|
|
|
|
|
|
|||||
Airport | MS | 1.75 | 7.04 | 9.85 | 14.73 |
MCRA | 1.64 | 7.54 | 9.66 | 15.15 | |
Proposed |
|
|
|
|
|
|
|||||
Street | MS | 2.77 | 6.75 |
|
|
MCRA | 2.34 | 7.40 | 9.88 | 14.36 | |
Proposed |
|
|
10.62 | 15.03 | |
|
|||||
Restaurant | MS | 0.31 | 4.48 | 9.40 | 14.74 |
MCRA | 0.27 | 5.47 | 9.20 | 15.24 | |
Proposed |
|
|
|
|
Objective evaluation and comparison of the Itakura-Saito Distortion Measure (ISDM).
Noise | Method | ISDM | |||
---|---|---|---|---|---|
0 (dB) | 5 (dB) | 10 (dB) | 15 (dB) | ||
White | MS | 1.20 | 0.87 | 0.60 | 0.42 |
MCRA | 0.92 | 0.44 | 0.55 | 0.37 | |
Proposed |
|
|
|
|
|
|
|||||
Car | MS | 0.15 | 0.16 | 0.02 | 0.01 |
MCRA | 0.21 | 0.05 | 0.02 | 0.01 | |
Proposed |
|
|
0.02 | 0.02 | |
|
|||||
Babble | MS | 0.15 | 0.06 | 0.02 | 0.01 |
MCRA | 0.12 | 0.02 | 0.01 | 0.01 | |
Proposed |
|
0.02 | 0.01 | 0.01 | |
|
|||||
Airport | MS | 0.19 | 0.06 | 0.02 | 0.02 |
MCRA | 0.14 | 0.04 | 0.02 | 0.01 | |
Proposed |
|
0.04 |
|
0.01 | |
|
|||||
Street | MS | 0.16 | 0.18 |
|
|
MCRA | 0.55 | 0.13 | 0.09 | 0.05 | |
Proposed | 0.17 |
|
0.08 | 0.04 | |
|
|||||
Restaurant | MS | 0.10 | 0.05 | 0.01 | 0.01 |
MCRA | 0.10 | 0.03 | 0.01 | 0.01 | |
Proposed | 0.10 |
|
0.01 | 0.01 |
The ISDM was shown to give a good correlation with subjective intelligibility measures specifically the diagnostic acceptability measure (DAM). This results in an objective test that can be used to produce a good meaningful result. This also results in a test that shows the distortion and noise reduction [
We presented a modified noise estimation and suppression algorithm that combined the nonlinear function and
The authors declare no competing interests.
This research was supported by NRF (2013R1A1A2012536).