Deep Learning-based Speech Enhancement Research Articles

A speech enhancement model based on noise component decomposition: Inspired by human cognitive behavior

In multimedia intelligent systems, speech enhancement is commonly employed to improve the quality of speech signals, making them clearer and more natural. Current deep learning-based speech enhancement models typically treat noise as a unified entity and aim to separate it from the target speech. In this paper, inspired by the cognitive behavior of the human brain when observing noisy speech spectrograms, we decompose the spectral energy of noise into regular and random components. We propose an auxiliary-model-based speech enhancement framework that better suppresses noise components closely resembling speech features. Firstly, we introduce a voiceprint segmentation network (VSnet) that partitions noisy speech into voiceprint and non-voiceprint regions. Subsequently, we present a noise reconstruction network (NRnet) that utilizes noise information from non-voiceprint regions to reconstruct and suppress the regular noise components within the voiceprint region. Finally, we construct a combination of a model dedicated to suppressing random components (RANnet) and a speech enhancement model (SEnet), and train them synchronously. By sharing encoder parameters, SEnet is compelled to reduce the extraction of regular noise features from the original noisy speech, contributing to improving speech quality generated through the decoder. Experimental results on public Voickbank-DEMAND and DNS-challenge 2020 datasets demonstrate that our approach achieves state-of-the-art performance.

Sub-convolutional U-Net with transformer attention network for end-to-end single-channel speech enhancement

Recent advancements in deep learning-based speech enhancement models have extensively used attention mechanisms to achieve state-of-the-art methods by demonstrating their effectiveness. This paper proposes a transformer attention network based sub-convolutional U-Net (TANSCUNet) for speech enhancement. Instead of adopting conventional RNNs and temporal convolutional networks for sequence modeling, we employ a novel transformer-based attention network between the sub-convolutional U-Net encoder and decoder for better feature learning. More specifically, it is composed of several adaptive time―frequency attention modules and an adaptive hierarchical attention module, aiming to capture long-term time-frequency dependencies and further aggregate hierarchical contextual information. Additionally, a sub-convolutional encoder-decoder model used different kernel sizes to extract multi-scale local and contextual features from the noisy speech. The experimental results show that the proposed model outperforms several state-of-the-art methods.

Speech Perception Improvement Algorithm Based on a Dual-Path Long Short-Term Memory Network.

Current deep learning-based speech enhancement methods focus on enhancing the time-frequency representation of the signal. However, conventional methods can lead to speech damage due to resolution mismatch problems that emphasize only specific information in the time or frequency domain. To address these challenges, this paper introduces a speech enhancement model designed with a dual-path structure that identifies key speech characteristics in both the time and time-frequency domains. Specifically, the time path aims to model semantic features hidden in the waveform, while the time-frequency path attempts to compensate for the spectral details via a spectral extension block. These two paths enhance temporal and spectral features via mask functions modeled as LSTM, respectively, offering a comprehensive approach to speech enhancement. Experimental results show that the proposed dual-path LSTM network consistently outperforms conventional single-domain speech enhancement methods in terms of speech quality and intelligibility.

Speech enhancement approach for body-conducted unvoiced speech based on Taylor–Boltzmann machines trained DNN

Communication becomes effectual while the verbal communication signal reaches the destination with profound characteristics. An application based on human voice attracts the researcher's attention. It leads the researcher to proceed towards the speech conversion for low voice humans having murmured voices as their attitude. The researchers introduced a technique for improving the body-conducted unvoiced speech for silent speech communication using Taylor-Boltzmann machine-trained optimized Deep Neural Network (DNN). The projected model includes two main stages: the training stage as well as recognition stage. In the training stage, the following functionalities are undergone: (i) pre-processing, (ii) feature extraction, and (iii) deep learning-based speech enhancement. Initially, the collected input Non-Audible Murmur (NAM) speech signal is subjected to the pre-processing stage, wherein the improved wiener filtering is employed for suppressing the noise in the input signal. From the pre-processed speech signal, the most relevant characteristics such as “spectral skewness, spectral chroma, spectral centroid and improved Mel Frequency Cepstral Coefficients (MFCC)” are extracted, and they are fused together. Then, using these fused features, the Taylor series with optimized DNN and Boltzmann Machine (Taylor-DNN-BM) model is trained. The final enhanced speech is acquired from the Taylor- optimized DNN-BM. The recognition phase encapsulates the feature extraction and deep learning-based speech enhancement stage. The input speech data enters into the feature extraction phase, wherein the characteristics such as “spectral skewness, spectral chroma, spectral centroid and improved (MFCC)” are retrieved. After that, these features are fused and fed as input to the Taylor-optimized DNN-BM model. From Taylor- optimized DNN-BM, the enhanced speech signal is acquired. In DNN, the weight function is fine-tuned using the newly projected Self Improved Flower Pollination Algorithm (SI-FPA) algorithm. This SI-FPA model is the conceptual improvement of the Flower Pollination Algorithm (FPA). Finally, the projected model is compared with the existing models to validate its efficiency in terms of speech enhancement.

Characterization of Deep Learning-Based Speech-Enhancement Techniques in Online Audio Processing Applications.

Deep learning-based speech-enhancement techniques have recently been an area of growing interest, since their impressive performance can potentially benefit a wide variety of digital voice communication systems. However, such performance has been evaluated mostly in offline audio-processing scenarios (i.e., feeding the model, in one go, a complete audio recording, which may extend several seconds). It is of significant interest to evaluate and characterize the current state-of-the-art in applications that process audio online (i.e., feeding the model a sequence of segments of audio data, concatenating the results at the output end). Although evaluations and comparisons between speech-enhancement techniques have been carried out before, as far as the author knows, the work presented here is the first that evaluates the performance of such techniques in relation to their online applicability. This means that this work measures how the output signal-to-interference ratio (as a separation metric), the response time, and memory usage (as online metrics) are impacted by the input length (the size of audio segments), in addition to the amount of noise, amount and number of interferences, and amount of reverberation. Three popular models were evaluated, given their availability on public repositories and online viability, MetricGAN+, Spectral Feature Mapping with Mimic Loss, and Demucs-Denoiser. The characterization was carried out using a systematic evaluation protocol based on the Speechbrain framework. Several intuitions are presented and discussed, and some recommendations for future work are proposed.

Dysarthric Speech Enhancement Based on Convolution Neural Network.

Generally, those patients with dysarthria utter a distorted sound and the restrained intelligibility of a speech for both human and machine. To enhance the intelligibility of dysarthric speech, we applied a deep learning-based speech enhancement (SE) system in this task. Conventional SE approaches are used for shrinking noise components from the noise-corrupted input, and thus improve the sound quality and intelligibility simultaneously. In this study, we are focusing on reconstructing the severely distorted signal from the dysarthric speech for improving intelligibility. The proposed SE system prepares a convolutional neural network (CNN) model in the training phase, which is then used to process the dysarthric speech in the testing phase. During training, paired dysarthric-normal speech utterances are required. We adopt a dynamic time warping technique to align the dysarthric-normal utter-ances. The gained training data are used to train a CNN - based SE model. The proposed SE system is evaluated on the Google automatic speech recognition (ASR) system and a subjective listening test. The results showed that the proposed method could notably enhance the recognition performance for more than 10% in each of ASR and human recognitions from the unprocessed dysarthric speech. Clinical Relevance- This study enhances the intelligibility and ASR accuracy from a dysarthria speech to more than 10.

A Waveform Mapping-Based Approach for Enhancement of Trunk Borers' Vibration Signals Using Deep Learning Model.

Simple SummaryTrunk-boring insects belong to one of the most destructive forest pests. Larvae in some groups are particularly difficult to detect since they make their living in trunks and no obvious sign can be found from outside. To deal with this problem, a new method is presented here, which is the embedding of a vibration probe into the tree trunk to pick up vibrations caused by larvae and the use of a model to distinguish whether the tree is infected. However, this procedure can experience severe interference from environmental noise, which is simultaneously picked up with vibrations. Thus, it is necessary to add a noise suppression process before discrimination. Previous examples have proved that the application of the analysis intended for sounds to boring vibrations is feasible. Therefore, we took advantage of deep learning-based speech enhancement and further improved it to develop a boring vibration enhancement model. The training data used in this research contains boring vibrations recorded within pieces of trunks and noise, which are common for trees’ living environment. The experimental results indicate that the enhancement procedure provided by our model substantially increases the accuracy of several well-known classification models, guaranteeing a more practical larvae detection.The larvae of some trunk-boring beetles barely leave traces on the outside of trunks when feeding within, rendering the detection of them rather difficult. One approach to solving this problem involves the use of a probe to pick up boring vibrations inside the trunk and distinguish larvae activity according to the vibrations. Clean boring vibration signals without noise are critical for accurate judgement. Unfortunately, these environments are filled with natural or artificial noise. To address this issue, we constructed a boring vibration enhancement model named VibDenoiser, which makes a significant contribution to this rarely studied domain. This model is built using the technology of deep learning-based speech enhancement. It consists of convolutional encoder and decoder layers with skip connections, and two layers of SRU++ for sequence modeling. The dataset constructed for study is made up of boring vibrations of Agrilus planipennis Fairmaire, 1888 (Coleoptera: Buprestidae) and environmental noise. Our VibDenoiser achieves an improvement of 18.57 in SNR, and it runs in real-time on a laptop CPU. The accuracy of the four classification models increased by a large margin using vibration clips enhanced by our model. The results demonstrate the great enhancement performance of our model, and the contribution of our work to better boring vibration detection.

Improved Speech Enhancement Considering Speech PSD Uncertainty

Speech enhancement based on statistical models has been studied for several decades. Recently, the speech enhancement adopting a speech power spectral density (PSD) uncertainty model has been proposed. This approach distinguishes the true speech PSD from its estimate and considers both as random variables. It incorporates a prior distribution of speech spectra and speech PSD estimators to derive the PSD uncertainty-aware counterpart to conventional clean speech estimators, which results in performance improvement. However, the speech PSD uncertainty model has not yet been adopted for parameter estimations such as <inline-formula><tex-math notation="LaTeX"><?TeX $\mathit{a posteriori}$?></tex-math></inline-formula> speech presence probability, noise PSD, and speech power spectra estimations in the speech enhancement framework. In this paper, we incorporate the speech PSD uncertainty model to all the components of the statistical model-based speech enhancement framework by deriving PSD uncertainty-aware counterparts to conventional parameter estimators. Specifically, we derive the <inline-formula><tex-math notation="LaTeX"><?TeX $\mathit{a posteriori}$?></tex-math></inline-formula> speech presence probability (SPP) where the likelihood function for each hypothesis is based on the speech PSD uncertainty. With this <inline-formula><tex-math notation="LaTeX"><?TeX $\mathit{a posteriori}$?></tex-math></inline-formula> SPP, a novel SPP-based noise PSD estimator is derived. Also, we derive the minimum mean-square error (MMSE) estimator for the power spectrum of the clean speech in the current frame under speech PSD uncertainty which is exploited to refine the speech PSD estimator. Finally, the refined speech PSD estimator is incorporated into the spectral gain function based on the speech PSD uncertainty model. The proposed approach showed improved noise PSD estimation performance in terms of the averaged logarithmic error distance, and improved speech enhancement performance in terms of the noise reduction, segmental signal-to-noise ratio, perceptual evaluation of speech quality (PESQ) scores and short-time objective intelligibility in our experiments. It also exhibited comparable performance with a real-time deep learning-based speech enhancement system in terms of the PESQ scores and composite measures for the VoiceBank-DEMAND dataset.

Monaural Speech Enhancement Using a Multi-Branch Temporal Convolutional Network

Deep learning has achieved substantial improvement on single-channel speech enhancement tasks. However, the performance of multi-layer perceptions (MLPs)-based methods is limited by the ability to capture the long-term effective history information. The recurrent neural networks (RNNs), e.g., long short-term memory (LSTM) model, are able to capture the long-term temporal dependencies, but come with the issues of the high latency and the complexity of training.To address these issues, the temporal convolutional network (TCN) was proposed to replace the RNNs in various sequence modeling tasks. In this paper we propose a novel TCN model that employs multi-branch structure, called multi-branch TCN (MB-TCN), for monaural speech enhancement.The MB-TCN exploits split-transform-aggregate design, which is expected to obtain strong representational power at a low computational complexity.Inspired by the TCN, the MB-TCN model incorporates one dimensional causal dilated CNN and residual learning to expand receptive fields for capturing long-term temporal contextual information.Our extensive experimental investigation suggests that the MB-TCNs outperform the residual long short-term memory networks (ResLSTMs), temporal convolutional networks (TCNs), and the CNN networks that employ dense aggregations in terms of speech intelligibility and quality, while providing superior parameter efficiency. Furthermore, our experimental results demonstrate that our proposed MB-TCN model is able to outperform multiple state-of-the-art deep learning-based speech enhancement methods in terms of five widely used objective metrics.

Deep Learning-Based Speech Enhancement With a Loss Trading Off the Speech Distortion and the Noise Residue for Cochlear Implants.

The cochlea plays a key role in the transmission from acoustic vibration to neural stimulation upon which the brain perceives the sound. A cochlear implant (CI) is an auditory prosthesis to replace the damaged cochlear hair cells to achieve acoustic-to-neural conversion. However, the CI is a very coarse bionic imitation of the normal cochlea. The highly resolved time-frequency-intensity information transmitted by the normal cochlea, which is vital to high-quality auditory perception such as speech perception in challenging environments, cannot be guaranteed by CIs. Although CI recipients with state-of-the-art commercial CI devices achieve good speech perception in quiet backgrounds, they usually suffer from poor speech perception in noisy environments. Therefore, noise suppression or speech enhancement (SE) is one of the most important technologies for CI. In this study, we introduce recent progress in deep learning (DL), mostly neural networks (NN)-based SE front ends to CI, and discuss how the hearing properties of the CI recipients could be utilized to optimize the DL-based SE. In particular, different loss functions are introduced to supervise the NN training, and a set of objective and subjective experiments is presented. Results verify that the CI recipients are more sensitive to the residual noise than the SE-induced speech distortion, which has been common knowledge in CI research. Furthermore, speech reception threshold (SRT) in noise tests demonstrates that the intelligibility of the denoised speech can be significantly improved when the NN is trained with a loss function bias to more noise suppression than that with equal attention on noise residue and speech distortion.

Target exaggeration for deep learning-based speech enhancement

Deep learning has been actively utilized for speech enhancement. However, deep learning-based speech enhancement usually produces over-smoothed speech, resulting in speech distortion and degraded intelligibility. In this paper, we propose the exaggeration of the training target so that the dynamic range of the enhanced speech becomes more similar to that of the clean speech. Target exaggeration can be implemented in two ways. The first approach is to exaggerate the target feature in the cost function of a deep learning-based speech enhancement system. This method can be implemented without additional parameters or computation, but can only be applied to schemes working in the time-frequency domain with the mean-square error cost function. The second approach is to introduce an additional deep neural network (DNN) that estimates the residual error in the output of a deep learning-based speech enhancement. This requires more computation, but can be applied even to time-domain approaches. To evaluate the performance of the proposed target exaggeration, it is applied to a feed-forward DNN- and long short-term memory (LSTM)-based speech enhancement scheme in the time-frequency domain, and the convolutional time-domain audio separation network (Conv-TasNet)-based speech enhancement scheme in the time domain. Experimental results showed that the proposed method improved the quality of speech produced by the deep learning-based speech enhancement system in terms of the perceptual evaluation of speech quality (PESQ) scores and outperformed other approaches, including global variance equalization and a perceptually optimized speech denoising autoencoder, to alleviate the over-smoothing problem.

Low-complexity artificial noise suppression methods for deep learning-based speech enhancement algorithms

Deep learning-based speech enhancement algorithms have shown their powerful ability in removing both stationary and non-stationary noise components from noisy speech observations. But they often introduce artificial residual noise, especially when the training target does not contain the phase information, e.g., ideal ratio mask, or the clean speech magnitude and its variations. It is well-known that once the power of the residual noise components exceeds the noise masking threshold of the human auditory system, the perceptual speech quality may degrade. One intuitive way is to further suppress the residual noise components by a postprocessing scheme. However, the highly non-stationary nature of this kind of residual noise makes the noise power spectral density (PSD) estimation a challenging problem. To solve this problem, the paper proposes three strategies to estimate the noise PSD frame by frame, and then the residual noise can be removed effectively by applying a gain function based on the decision-directed approach. The objective measurement results show that the proposed postfiltering strategies outperform the conventional postfilter in terms of segmental signal-to-noise ratio (SNR) as well as speech quality improvement. Moreover, the AB subjective listening test shows that the preference percentages of the proposed strategies are over 60%.

Towards Model Compression for Deep Learning Based Speech Enhancement.

The use of deep neural networks (DNNs) has dramatically elevated the performance of speech enhancement over the last decade. However, to achieve strong enhancement performance typically requires a large DNN, which is both memory and computation consuming, making it difficult to deploy such speech enhancement systems on devices with limited hardware resources or in applications with strict latency requirements. In this study, we propose two compression pipelines to reduce the model size for DNN-based speech enhancement, which incorporates three different techniques: sparse regularization, iterative pruning and clustering-based quantization. We systematically investigate these techniques and evaluate the proposed compression pipelines. Experimental results demonstrate that our approach reduces the sizes of four different models by large margins without significantly sacrificing their enhancement performance. In addition, we find that the proposed approach performs well on speaker separation, which further demonstrates the effectiveness of the approach for compressing speech separation models.

Towards More Efficient DNN-Based Speech Enhancement Using Quantized Correlation Mask

Many studies on deep learning-based speech enhancement (SE) utilizing the computational auditory scene analysis method typically employs the ideal binary mask or the ideal ratio mask to reconstruct the enhanced speech signal. However, many SE applications in real scenarios demand a desirable balance between denoising capability and computational cost. In this study, first, an improvement over the ideal ratio mask to attain more superior SE performance is proposed through introducing an efficient adaptive correlation-based factor for adjusting the ratio mask. The proposed method exploits the correlation coefficients among the noisy speech, noise and clean speech to effectively re-distribute the power ratio of the speech and noise during the ratio mask construction phase. Second, to make the supervised SE system more computationally-efficient, quantization techniques are considered to reduce the number of bits needed to represent floating numbers, leading to a more compact SE model. The proposed quantized correlation mask is utilized in conjunction with a 4-layer deep neural network (DNN-QCM) comprising dropout regulation, pre-training and noise-aware training to derive a robust and high-order mapping in enhancement, and to improve generalization capability in unseen conditions. Results show that the quantized correlation mask outperforms the conventional ratio mask representation and the other SE algorithms used for comparison. When compared to a DNN with ideal ratio mask as its learning targets, the DNN-QCM provided an improvement of approximately 6.5% in the short-time objective intelligibility score and 11.0% in the perceptual evaluation of speech quality score. The introduction of the quantization method can reduce the neural network weights to a 5-bit representation from a 32-bit, while effectively suppressing stationary and non-stationary noise. Timing analyses also show that with the techniques incorporated in the proposed DNN-QCM system to increase its compactness, the training and inference time can be reduced by 15.7% and 10.5%, respectively.

An Experimental Analysis of Deep Learning Architectures for Supervised Speech Enhancement

Recent speech enhancement research has shown that deep learning techniques are very effective in removing background noise. Many deep neural networks are being proposed, showing promising results for improving overall speech perception. The Deep Multilayer Perceptron, Convolutional Neural Networks, and the Denoising Autoencoder are well-established architectures for speech enhancement; however, choosing between different deep learning models has been mainly empirical. Consequently, a comparative analysis is needed between these three architecture types in order to show the factors affecting their performance. In this paper, this analysis is presented by comparing seven deep learning models that belong to these three categories. The comparison includes evaluating the performance in terms of the overall quality of the output speech using five objective evaluation metrics and a subjective evaluation with 23 listeners; the ability to deal with challenging noise conditions; generalization ability; complexity; and, processing time. Further analysis is then provided while using two different approaches. The first approach investigates how the performance is affected by changing network hyperparameters and the structure of the data, including the Lombard effect. While the second approach interprets the results by visualizing the spectrogram of the output layer of all the investigated models, and the spectrograms of the hidden layers of the convolutional neural network architecture. Finally, a general evaluation is performed for supervised deep learning-based speech enhancement while using SWOC analysis, to discuss the technique’s Strengths, Weaknesses, Opportunities, and Challenges. The results of this paper contribute to the understanding of how different deep neural networks perform the speech enhancement task, highlight the strengths and weaknesses of each architecture, and provide recommendations for achieving better performance. This work facilitates the development of better deep neural networks for speech enhancement in the future.

RETRACTED: Speech enhancement method using deep learning approach for hearing-impaired listeners.

A deep learning-based speech enhancement method is proposed to aid hearing-impaired listeners by improving speech intelligibility. The algorithm decomposes the noisy speech signal into frames (as features). Subsequently, a deep convolutional neural network is fed with decomposed noisy speech signal frames to produce frequency channel estimation. However, a higher signal-to-noise ratio information is contained in produced frequency channel estimation. Using this estimate, speech-dominated cochlear implant channels are taken to produce electrical stimulation. This process is the same as that of the conventional n-of-m cochlear implant coding strategies. To determine the speech-in-noise performance of 12 cochlear implant users, the fan and music sound applied are considered as background noises. Performance of the proposed algorithm is evaluated by considering these background noises. Low processing delay and reliable architecture are the best characteristics of the deep learning-based speech enhancement algorithm; hence, this can be suitably applied for all applications of hearing devices. Experimental results demonstrate that deep convolutional neural network approach appeared more promising than conventional approaches.

On Cross-Corpus Generalization of Deep Learning Based Speech Enhancement.

In recent years, supervised approaches using deep neural networks (DNNs) have become the mainstream for speech enhancement. It has been established that DNNs generalize well to untrained noises and speakers if trained using a large number of noises and speakers. However, we find that DNNs fail to generalize to new speech corpora in low signal-to-noise ratio (SNR) conditions. In this work, we establish that the lack of generalization is mainly due to the channel mismatch, i.e. different recording conditions between the trained and untrained corpus. Additionally, we observe that traditional channel normalization techniques are not effective in improving cross-corpus generalization. Further, we evaluate publicly available datasets that are promising for generalization. We find one particular corpus to be significantly better than others. Finally, we find that using a smaller frame shift in short-time processing of speech can significantly improve cross-corpus generalization. The proposed techniques to address cross-corpus generalization include channel normalization, better training corpus, and smaller frame shift in short-time Fourier transform (STFT). These techniques together improve the objective intelligibility and quality scores on untrained corpora significantly.

Increasing Compactness of Deep Learning Based Speech Enhancement Models With Parameter Pruning and Quantization Techniques

Most recent studies on deep learning based speech enhancement (SE) focused on improving denoising performance. However, successful SE applications require striking a desirable balance between denoising performance and computational cost in real scenarios. In this study, we propose a novel parameter pruning (PP) technique, which removes redundant channels in a neural network. In addition, a parameter quantization (PQ) technique was applied to reduce the size of a neural network by representing weights with fewer cluster centroids. Because the techniques are derived based on different concepts, the PP and PQ can be integrated to provide even more compact SE models. The experimental results show that the PP and PQ techniques produce a compacted SE model with a size of only 10.03% compared to that of the original model, resulting in minor performance losses of 1.43% (from 0.70 to 0.69) for STOI and 3.24% (from 1.85 to 1.79) for PESQ. The promising results suggest that the PP and PQ techniques can be used in a SE system in devices with limited storage and computation resources.

Using Generalized Gaussian Distributions to Improve Regression Error Modeling for Deep Learning-Based Speech Enhancement

From a statistical perspective, the conventional minimum mean squared error (MMSE) criterion can be considered as the maximum likelihood (ML) solution under an assumed homoscedastic Gaussian error model. However, in this paper, a statistical analysis reveals the super-Gaussian and heteroscedastic properties of the prediction errors in nonlinear regression deep neural network (DNN)-based speech enhancement when estimating clean log-power spectral (LPS) components at DNN outputs with noisy LPS features in DNN input vectors. Accordingly, we propose treating all dimensions of the prediction error vector as statistically independent random variables and model them with generalized Gaussian distributions (GGDs). Then, the objective function with the GGD error model is derived according to the ML criterion. Experiments on the TIMIT corpus corrupted by simulated additive noises show consistent improvements of our proposed DNN framework over the conventional DNN framework in terms of various objective quality measures under 14 unseen noise types evaluated and at various signal-to-noise ratio levels. Furthermore, the ML optimization objective with GGD outperforms the conventional MMSE criterion, achieving improved generalization and robustness.

Disentangled Feature Learning for Noise-Invariant Speech Enhancement

Most of the recently proposed deep learning-based speech enhancement techniques have focused on designing the neural network architectures as a black box. However, it is often beneficial to understand what kinds of hidden representations the model has learned. Since the real-world speech data are drawn from a generative process involving multiple entangled factors, disentangling the speech factor can encourage the trained model to result in better performance for speech enhancement. With the recent success in learning disentangled representation using neural networks, we explore a framework for disentangling speech and noise, which has not been exploited in the conventional speech enhancement algorithms. In this work, we propose a novel noise-invariant speech enhancement method which manipulates the latent features to distinguish between the speech and noise features in the intermediate layers using adversarial training scheme. To compare the performance of the proposed method with other conventional algorithms, we conducted experiments in both the matched and mismatched noise conditions using TIMIT and TSPspeech datasets. Experimental results show that our model successfully disentangles the speech and noise latent features. Consequently, the proposed model not only achieves better enhancement performance but also offers more robust noise-invariant property than the conventional speech enhancement techniques.