Design of Speech Denoising Algorithm Based on Wavelet Threshold Function and PSO

At present, there are many shortcomings in the discontinuity of wavelet threshold function and the constant threshold of different decomposition layers and the constant error it produced. The amplitude-frequency characteristics of wavelet filters are studied and analyzed by mathematical modeling. An improved wavelet threshold function with adjustable parameters is proposed. Particle swarm optimization (PSO) algorithm is used to find the optimal parameters of the improved threshold function in a background noise environment. The improved wavelet threshold function is combined with Bayesian threshold method to obtain the threshold based on Bayesian criterion, which makes the threshold adaptive in different layers and overcomes the shortcomings of fixed threshold. Finally, the speech signal with optimal wavelet coefficients is obtained after reconstruction. Compared with the traditional threshold function, Simulation results show that the improved threshold function achieves precise notch denoising, effectively retains the singularity and eigenvalues of the signal, and reduces the signal distortion.

Cascaded Convolutional Neural Network Architecture for Speech Emotion Recognition in Noisy Conditions

Convolutional neural networks (CNNs) are a state-of-the-art technique for speech emotion recognition. However, CNNs have mostly been applied to noise-free emotional speech data, and limited evidence is available for their applicability in emotional speech denoising. In this study, a cascaded denoising CNN (DnCNN)–CNN architecture is proposed to classify emotions from Korean and German speech in noisy conditions. The proposed architecture consists of two stages. In the first stage, the DnCNN exploits the concept of residual learning to perform denoising; in the second stage, the CNN performs the classification. The classification results for real datasets show that the DnCNN–CNN outperforms the baseline CNN in overall accuracy for both languages. For Korean speech, the DnCNN–CNN achieves an accuracy of 95.8%, whereas the accuracy of the CNN is marginally lower (93.6%). For German speech, the DnCNN–CNN has an overall accuracy of 59.3–76.6%, whereas the CNN has an overall accuracy of 39.4–58.1%. These results demonstrate the feasibility of applying the DnCNN with residual learning to speech denoising and the effectiveness of the CNN-based approach in speech emotion recognition. Our findings provide new insights into speech emotion recognition in adverse conditions and have implications for language-universal speech emotion recognition.

A Study on Effectiveness of Deep Neural Networks for Speech Signal Enhancement in Comparison with Wiener Filtering Technique

We have compared two Neural network models with Wiener filtering technique for Speech signal enhancement. Our paper intends to suggest the best method suitable for speech denoising and quality enhancement. We have utilized MATLAB software with most advanced toolboxes for building the models. For comparing our models, we computed PSNR and SNR values.

A Study on Effectiveness of Deep Neural Networks for Speech Signal Enhancement in Comparison with Wiener Filtering Technique

We have compared two Neural network models with Wiener filtering technique for Speech signal enhancement. Our paper intends to suggest the best method suitable for speech denoising and quality enhancement. We have utilized MATLAB software with most advanced toolboxes for building the models. For comparing our models, we computed PSNR and SNR values.