Results 41 to 50 of about 3,717 (149)
A Comparative Study of Speech Coding Techniques for Electro Larynx Speech Production
Speech coding is a method of earning a tight speech signals representation for efficient storage and efficient transmission over band limited wired or wireless channels.
Zinah J. Mohammed +1 more
doaj +1 more source
The study collects typical music samples and multiple types of noise data, preprocesses the signals using time‐frequency domain feature extraction methods, and on this basis designs a deep neural network structure that integrates convolutional neural networks and perception optimization mechanisms.
Chao Jiang, Junfang Liang
wiley +1 more source
Robust Dysarthric Speech Recognition with GAN Enhancement and LLM Correction
This study tackles dysarthric speech recognition by combining generative adversarial network (GAN)‐generated synthetic data with large language model (LLM)‐based error correction. The approach integrates three key elements: an improved CycleGAN to generate synthetic dysarthric speech for data augmentation, a multimodal automatic speech recognition core
Yibo He +3 more
wiley +1 more source
Does a PESQNet (Loss) Require a Clean Reference Input? The Original PESQ Does, But ACR Listening Tests Don\u27t [PDF]
Perceptual evaluation of speech quality (PESQ) requires a clean speech reference as input, but predicts the results from (reference-free) absolute category rating (ACR) tests.
Strake, Maximilian +2 more
core
This workflow presents a complete pipeline for audio data processing, beginning with format conversion, channel adjustments, and cleaning, followed by enhancement and visualization techniques. It further applies signal separation using FastICA, postprocessing, and evaluation metrics (SDR, SIR, SAR) to improve audio analysis and support future research ...
Md. Razu Ahmed +3 more
wiley +1 more source
Bi‐DecoNet: Compact Dual‐Decoder Network for Robust Speech Enhancement Under Industrial Noise
Bi‐DecoNet is a compact dual‐decoder fusion network for robust speech enhancement under industrial noise. It integrates masking and spectral mapping with a lightweight time‐frequency adaptive transformer, a dilated DenseNet, and an attentional feature fusion module to effectively enhance noisy speech. Experiments on VoiceBank+DEMAND and TIMIT show that
Zhongchao Guan +3 more
wiley +1 more source
Acoustic echo cancellation (AEC) in actual communication systems is challenging due to highly correlated inputs, impulsive disturbances, and computational limitations. The traditional Affine Projection Algorithm (APA) is better than the Normalised Least Mean Square method, although it involves matrix inversion complexity and is susceptible to outliers (
Gagandeep Singh +5 more
wiley +1 more source
DeepSC‐S‐MECA: A Hybrid Mamba‐ECA Framework for Robust Semantic Speech Communication
This paper proposes DeepSC‐S‐MECA, a semantic communication model that integrates a Mamba Transformer and ECA‐ResNet to effectively capture temporal–contextual dependencies and emphasize informative speech features while maintaining low computational complexity.
Yerin Yeo, Junghyun Kim
wiley +1 more source
Efficient Harmonic Network With Cayley Polar Complex Mask for Single‐Channel Speech Enhancement
This article presents a spectral domain speech enhancement network that combines efficient harmonic modeling with a Cayley polar mask. The proposed harmonic neighbor injection module captures half‐, same‐, and double‐frequency relations with linear complexity, while the Cayley polar mask provides stable two‐parameter magnitude and phase correction ...
Junkang Yang, Ruibiao Zhu
wiley +1 more source
Employing Real Training Data for Deep Noise Suppression [PDF]
Most deep noise suppression (DNS) models are trained with reference-based losses requiring access to clean speech. However, sometimes an additive microphone model is insufficient for real-world applications. Accordingly, ways to use real training data in
Pirklbauer, Jan +3 more
core +1 more source

