Speech Dereverberation Based on Integrated Deep and Ensemble Learning Algorithm

Abstract

Reverberation, which is generally caused by sound reflections from walls, ceilings, and floors, can result in severe performance degradation of acoustic applications. Due to a complicated combination of attenuation and time-delay effects, the reverberation property is difficult to characterize, and it remains a challenging task to effectively retrieve the anechoic speech signals from reverberation ones. In the present study, we proposed a novel integrated deep and ensemble learning algorithm (IDEA) for speech dereverberation. The IDEA consists of offline and online phases. In the offline phase, we train multiple dereverberation models, each aiming to precisely dereverb speech signals in a particular acoustic environment; then a unified fusion function is estimated that aims to integrate the information of multiple dereverberation models. In the online phase, an input utterance is first processed by each of the dereverberation models. The outputs of all models are integrated accordingly to generate the final anechoic signal. We evaluated the IDEA on designed acoustic environments, including both matched and mismatched conditions of the training and testing data. Experimental results confirm that the proposed IDEA outperforms single deep-neural-network-based dereverberation model with the same model architecture and training data.

Speech Enhancement Based on Denoising Autoencoder With Multi-Branched Encoders

“…As reported in [48], the DDAE with a highway structure, termed HDDAE, is more robust than the conventional DDAE, and thus we will focus on the HDDAE model in this study. This HDDAE model includes a link that copies the front hidden layers to the later hidden layers to incorporate low-level information into the supervised stage.…”

“…First, we followed the setup of the best model DAEME-USAT (W D) (12) and adopted HDDAE in addition to BLSTM as the architecture of a component model. The BLSTMbased component model consisted of two layers, with 300 memory cells in each layer, and the HDDAE-based component model consisted of five hidden layers, each containing 2,048 neurons, and one highway layer [48]. Figs.…”

“…Examples include acoustic modeling [45], image classification [46], and bio-medical analysis [47]. Ensemble learning algorithms have also been used for speech signal processing, e.g., speech dereverberation [48] and SE. Lu et al investigated ensemble learning using unsupervised partitioning of training data [49].…”

Ensemble Hierarchical Extreme Learning Machine for Speech Dereverberation

“…During the online stage, the dereverberated magnitude and the phase spectrums of the original signal are used to reconstruct the waveform. In [43], a DDAE-based integrated deep and ensemble learning algorithm (IDEA) was proposed to effectively reduce the reverberation artifacts. In addition to DDAEbased ensemble models, the authors in [43] utilized a highway strategy and proposed a highway-DDAE (DDAE(Hwy)) framework to further improve the speech dereverberation performance.…”

“…In [43], a DDAE-based integrated deep and ensemble learning algorithm (IDEA) was proposed to effectively reduce the reverberation artifacts. In addition to DDAEbased ensemble models, the authors in [43] utilized a highway strategy and proposed a highway-DDAE (DDAE(Hwy)) framework to further improve the speech dereverberation performance. In this study, we extend the IDEA framework by replacing the HDDAE blocks with more effective residual-DDAE (DDAE(Res)) blocks to prepare ensemble models.…”