Levent Arslan, Alan McCree, and Vishu Viswanathan, "New methods for adaptive noise suppression," in Proceedings of the IEEE International Conference on Acoustics, Speech, and Signal Processing,May 1995, vol. 1, pp. 812--815.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....the noisy spectrum as: 0 # Y i (k) else (7) where the spectral floor parameter was set to # =0.002. 3. IMPLEMENTATION Speech was Hamming windowed using a 20 ms window and a 10 ms overlap between frames. The Fast Fourier Transform (FFT) of the windowed speech was smoothed as per [8] and a weighted spectral average [9] 10] is taken over preceding and succeeding frames of data as: Y j (k) M X l=#M W l Y j#l (k) 8) where j is the frame index. The number of frames, M, was limited to 2 to prevent spectral smearing. The filter weights W l were empirically determined and ....

L. Arslan, A. McCree and V. Viswanathan, "New methods for adaptive noise suppression," ICASSP, vol.1, pp. 812-815, May 1995.

....background noise is a key problem for noise reduction algorithms. Generally, noise is assumed to be additive and stationary with respect to speech. We use a a simple method of adaptive noise estimation to get a continuous noise estimate in order to avoid the use of a VAD (Voice Activity Detector) [8]. The noise estimate is continually updated but is allowed to increase much more slowly than it is allowed to decrease. Thus the noise estimate will increase only slowly during speech intervals and collapse quickly back during speech gaps. The noise estimate update (omitting the slow increase ....

L. Arslan, A. McCree, and V. Viswanathan, "New methods for adaptive noise suppression," in Proc. ICASSP, Detroit, USA, 1995, pp. 812--815.

....in speech pauses. It also allows implementation of the single input SE algorithm. However, it is known that the VAD needed for the pause detection operates properly only under high SNR conditions. Low SNR conditions do not allow to use the VAD. Some SS algorithms without the VAD are presented in [2,3]. However, it was found that the method [2] does not operate properly in low SNR conditions and the method [3] has some performance degradation in pause less speech. Method [1] was found as the most suitable for these conditions. 2. MODIFIED SPECTRAL SUBTRACTION ALGORITHM The key idea of ESS ....

....of the single input SE algorithm. However, it is known that the VAD needed for the pause detection operates properly only under high SNR conditions. Low SNR conditions do not allow to use the VAD. Some SS algorithms without the VAD are presented in [2,3] However, it was found that the method [2] does not operate properly in low SNR conditions and the method [3] has some performance degradation in pause less speech. Method [1] was found as the most suitable for these conditions. 2. MODIFIED SPECTRAL SUBTRACTION ALGORITHM The key idea of ESS [1] is to estimate noise rather than speech. ....

[Article contains additional citation context not shown here]

L. Arslan, A. McCree, V. Viswanathan, New method for Adaptive Noise Suppression.Proceedings of International Conference on Acoustic, Speech and Signal Processing (ICASSP95), 1995, pages 812815.

....front end speech enhancement algorithms[1] A number of speech enhancement algorithms have been proposed in the past. A survey can be found in [2] as well as an overview of statistical based approaches in [3] Several enhancement approaches have been proposed using improved SNR characterization[4], linear and nonlinear spectral subtraction[5, 6] and Wiener filtering[7] Traditional speech enhancement methods are based on optimizing mathematical criteria, which in general are not always well correlated with speech perception. Several recent methods have also considered auditory processing ....

L. Arslan, A. McCree, and V. Viswanathan, "New Methods for Adaptive Noise Suppression," Proc. 1995 IEEE ICASSP, pp. 812-815.

.... processing in the short time spectral domain (Boll, 1979; Weiss and Aschkenasy, 1983; Berouti et al. 1979; McAulay and Malpass, 1980) while later approaches were formulated based upon all pole modeling of the speech production system (Lim and Oppenheim, 1978; Hansen and Clements, 1991; Hansen and Arslan, 1995). Modern approaches, however, have investigated the application of hidden Markov models into an enhancement framework (Ephraim, 1992a; Ephraim 1992b) as well as the inclusion of auditory based enhancement processing (Cheng and 2 O Shaughnessy, 1991; Nandkumar and Hansen, 1995; Azirani et al. ....

....artifacts may become just as overwhelming as the original degradation itself and further lead to losses in intelligibility. Recent studies have demonstrated that improved enhancement can result when the phonetic inventory of the signal is considered during enhancement processing. In Hansen and Arslan (1995), the constrained iterative enhancement algorithm formulated by Hansen and Clements (1987, 1991) was shown to provide improved enhancement by adaptation of the terminating iteration using broad phoneme classifications. In this scheme, a noisy speech trained Markov model based phoneme classifier ....

[Article contains additional citation context not shown here]

L. Arslan, A. McCree, and V. Viswanathan (1995), "New Methods for Adaptive Noise Suppression," Proc. 1995 IEEE Inter. Conf. on Acoustics, Speech, and Signal Processing, pp. 812-815.

....found in texts on the subject, such as those by Lim [29] Deller et al. [30] and Furui and Sondhi [31] Spectral modification techniques were among the earliest and simplest of the enhancement methods. These methods include Boll s original spectral subtraction method [32] and its many variations [33, 34, 35, 36] as well as extensions to generalized spectra by Ephraim and Van Trees [37] and multi resolution analysis by Seok and Bae [38] Another approach involves stochastic modeling of the speech and noise signals followed by an appropriately derived filter. Lim and Oppenheim [39] proposed modeling the ....

L. Arslan, A. McCree, and V. Viswanathan, "New methods for adaptive noise suppression," in Proceedings of ICASSP95, pp. 812--815, IEEE, 1995.

....Wiener filtering alone. 1. INTRODUCTION The handset equipment for telephones in cars is a restriction and a potential risk for the driver. Only hands free devices can overcome this problem. Two different approaches for hands free devices can be pursued. The first one uses only one microphone [1, 2], whereas the second one is a multichannel approach [3, 4] The most often used single sensor method is spectral subtraction. However, this method introduces various other problems, as e.g. musical tones. Using smoothed versions of the spectral subtraction in order to avoid these tones leads to ....

L. Arslan, A. McCree, and V. Viswanathan, "New methods for adaptive noise suppression," in Proc. IEEE Int. Conference Acoustic, Speech and Signal Processing, ICASSP--95, (Detroit, Michigan), pp. 812--815, Mai 1995.

....noise 1 Musical noise is so termed because it consists of short tones at random frequencies. below the threshold of audibility [15] smoothing techniques, which represent the spectral subtraction process or Wiener filtering as a time varying filter which is smoothed both in time and frequency [2]; and methods which attempt to find optimal spectral subtraction parameters [7, 3] This work utilizes the multi resolution sinusoidal transform (MRST) to obtain signal parameters which are then processed using a modified Wiener filter algorithm. The resulting parameters are then conditioned to ....

.... to 32 points representing a special noise spectrum, N( l ; m) 6] N( l ; m 1) is then updated by averaging it with X( l ; m) but only allowing a slight decrease and an even smaller increase from N( l ; m) This causes N( l ; m) to tend toward an estimate of the noise spectrum [2]. Speech is assumed present if: 1. the signal exceeded a minimum energy level by 10 dB, or 2. P 31 l=0 fi fi fi X( l ; m) Gamma N( l ; m) fi fi fi exceeds the minimum RMS level by 8 dB. The noise spectrum, N( m) is estimated by averaging the signal spectra over time when no speech is ....

Levent Arslan, Alan McCree, and Vishu Viswanathan. New methods for adaptive noise suppression. In Proceedings of the International Conference on Acoustics, Speech, and Signal Processing, volume 1, pages 812--815, May 1995.

....(VAD) does not work well, as the VAD assumes that the noise is stationary during speech intervals. Furthermore, there are no robust one channel VADs up to now. To overcome this problem, several one channel noise estimation techniques without VAD have been introduced during the last few years [2, 3, 4, 5, 6]. Another way of handling this problem is to use a two channel approach to estimate the noise [1, 7] In the first part of our work we will give a short summary of the existing techniques. Secondly, a new algorithm is introduced and we show that this algorithm theoretically can estimate the noise ....

....is known. Finally, we will compare all algorithms in terms of their abilities to estimate the noise of a mixed signal, consisting of slowly amplitude modulated noise added to clean speech. 2. ALGORITHMS 2.1. One Channel Algorithms ffl Voice Activity Detector (VAD) ffl Direct Estimation (DE) [3] ffl Modified Direct Estimation (MDE) 3] ffl Threshold Direct Estimation (TDE) 5] ffl Histogram Technique (HT) 5] ffl Minimum Statistics, Martin (MSM) 2] ffl Minimum Statistics, Doblinger (MSD) 4] ffl Iterative Wiener Filter (IWF) 6] All one channel noise estimation techniques use ....

[Article contains additional citation context not shown here]

L. Arslan, A. McCree, and V. Viswanathan, "New Methods for Adaptive Noise Suppression, " in Proc. IEEE Int. Conference Acoustic, Speech and Signal Processing, ICASSP--95, (Detroit, Michigan), pp. 812--815, Mai 1995.

No context found.

L.M. Arslan, A. McCree, and V. Viswanathan (1995). "New Methods for Adaptive Noise Suppression". In Proc. IEEE Int. Conf. Acoust., Speech, Signal Processing, volume 1, pp. 812--815, Detroit, USA.

....lower order LSFs, and for unvoiced segments, higher order LSFs are weighted more by an exponential weighting factor. Based on the distances from each codebook entry, an approximate line spectral frequency vector can be expressed as a weighted sum of the source codebook line spectral frequencies [2]: v i = e Gammafl d i P L l=1 e Gammafl d l i = 1; L wk = P L i=1 v i S ik k = 1; P (3) where the value of fl for each frame is found by an incremental search with the criterion of minimizing the perceptual weighted distance between the approximated LSF vector w and ....

L.M. Arslan, A. McCree, and V. Viswanathan. "New Methods for Adaptive Noise Suppression". In Proc. IEEE Int. Conf. Acoust., Speech, Signal Processing, volume 1, pages 812--815, Detroit, USA, May 1995.

No context found.

L.M. Arslan, A. McCree, and V. Viswanathan (1995). "New Methods for Adaptive Noise Suppression". In Proc. IEEE Int. Conf. Acoust., Speech, Signal Processing, Vol. 1, pp. 812-- 815, Detroit, USA.

....interfering noise. The channel distortion normally possesses nonstationary statistics, and can contain correlated interference (e.g. another speaker s voice) However, most models developed for single channel speech enhancement systems assume that background noise is stationary and or uncorrelated[1, 2, 4, 7]. Although the fundamental principles behind these enhancement methods are well defined, in practice, the limitations set by their assumptions play a major role in their performance across actual distortions. The reason for this is that they rely on a good estimate of the noise characteristics, ....

L.M. Arslan, A. McCree, and V. Viswanathan. "New Methods for Adaptive Noise Suppression". In Proc. IEEE Int. Conf. Acoust., Speech, Signal Processing, volume 1, pages 812--815, Detroit, USA, May 1995.

No context found.

Levent Arslan, Alan McCree, and Vishu Viswanathan, "New methods for adaptive noise suppression," in Proceedings of the IEEE International Conference on Acoustics, Speech, and Signal Processing,May 1995, vol. 1, pp. 812--815.

No context found.

L. Arslan, A. McCree, and V. Viswanathan. New methods for adaptive noise suppression. In Proc. ICASSP'95, pages 812--815, Detroit, Michigan, May 1995.

No context found.

L. Arslan, A. McCree, and V. Viswanathan, "New Methods for Adaptive Noise Suppression," Proc. 1995 IEEE Inter. Conf. on Acoustics, Speech, and Signal Processing, pp. 812-815.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC