Abstract:This paper addresses the problem of single microphone speech enhancement in noisy environments. Common short-time noise reduction techniques introduce harmonic distortion in enhanced speech because of the non reliability of estimators for small signalto-noise ratios. We propose a new method called Harmonic Regeneration Noise Reduction technique which solves this problem. A fully harmonic signal is calculated based on the distorted signal using a non-linearity to regenerate harmonics in an efficient way. This a… Show more
“…A several one-, two-and multichannel sensors techniques are proposed to deal with this problem. For example in [14][15][16][17], several single and two-sensor techniques are proposed to correct these distortions [18][19][20]. More advanced techniques are then proposed recently in [21,22].…”
“…A several one-, two-and multichannel sensors techniques are proposed to deal with this problem. For example in [14][15][16][17], several single and two-sensor techniques are proposed to correct these distortions [18][19][20]. More advanced techniques are then proposed recently in [21,22].…”
“…To improve the noise-robustness of an ASR system trained using clean data, techniques such as speech enhancement [1][2][3][4], noise-robust feature extraction feature enhancement [5] [6], and model-based noise adaptation [7] can be applied. Most of these techniques require a reliable voice activity detector (VAD) to identify non-speech segments for noise estimation.…”
Section: Introductionmentioning
confidence: 99%
“…Nonlinear spectral subtraction (NSS) in [3] defined a smaller subtraction factor at harmonic peaks which resulted in higher ASR accuracy at low SNRs. In [4], regeneration of harmonic structures improved the quality of denoised speech. A harmonic model is used in a generalized LRT for robust voiced/unvoiced detection in [8], and a periodic-toaperiodic component ratio used for speech/non-speech detection in [9] showed promising results with aperiodic noise interferences.…”
This paper proposes a new statistical model-based likelihood ratio test (LRT) VAD to obtain reliable speech / non-speech decisions. In the proposed method, the likelihood ratio (LR) is calculated differently for voiced frames, as opposed to unvoiced frames: only DFT bins containing harmonic spectral peaks are selected for LR computation. To evaluate the new VAD's effectiveness in improving the noiserobustness of ASR, its decisions are applied to preprocessing techniques such as non-linear spectral subtraction, minimum mean square error short-time spectral amplitude estimator, and frame dropping. From the ASR experiments conducted on the Aurora2 database, the proposed harmonic frequency-based LRTs give better results than conventional LRT-based VADs and the standard G.729B and ETSI AMR VADs.
“…The distorted signal is processed to create a fully harmonic signal where all the missing harmonics are regenerated. Hence, this method is called as Harmonic Regeneration Noise Reduction (HRNR) method and it was presented in [3], [5], [8], [9]. Two-step noise reduction (TSNR), to refine the estimation of the a priori SNR which get rid of the drawbacks of the DD approach while maintaining its advantage, i.e.…”
The performance of a noisy speech enhancement algorithm depends on the estimation of the priori signal-to-noise ratio (SNR). The most commonly used approach to estimate the priori SNR parameter uses Decision-Directed (DD), TwoStep Noise Reduction (TSNR) and Harmonic Regeneration Noise Reduction (HRNR) method. Two-step noise reduction (TSNR) method eliminate this problem decision-directed method. Common short-time noise reduction techniques introduce harmonic distortion in enhanced speech because of the non-reliability of estimators for small signal to-noise ratios. A simple but effective harmonic regeneration method called Harmonic Regeneration Noise Reduction (HRNR) is used to overcome this problem.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.