Noise Suppression in Speech Signals using Artificial Intelligence

Zıvalıoğlu C. C., Akay O.

INTERNATIONAL CONFERENCE on ELECTRICAL and ELECTRONICS ENGINEERING (ELECO 2025), Bursa, Turkey, 27 - 29 November 2025, pp.1-5, (Full Text)

  • Publication Type: Conference Paper / Full Text
  • City: Bursa
  • Country: Turkey
  • Page Numbers: pp.1-5
  • Dokuz Eylül University Affiliated: Yes

Abstract

This study presents a hybrid approach combining deep learning-based noise classification and subspace projection for environmental noise suppression in speech signals. Convolutional Neural Networks (CNNs), Long Short-Term Memory (LSTM) networks, and a CNN-LSTM hybrid model are employed for noise type classification. Prediction smoothing is applied to improve robustness in silent and lowSNR regions of speech. The proposed system achieves noticeable improvements in terms of signal-to-noise ratio (SNR), short-time objective intelligibility (STOI), and perceptual evaluation of speech quality (PESQ) under stationary and nonstationary noise scenarios.