31, August 2026

CNN and DenseNet201-Based Speech Enhancement for Hearing-Assistive Applications

VOLUME-12; ISSUE-8; AUG-2026

Author(s): 1. Dr. S. Swarnalatha, 2. D. Pooja

Authors Affiliations:

1 Professor, Department of Electronics and Communication Engineering, S V University College of Engineering, Tirupati, Andhra Pradesh, India

2 Student, M.Tech, Department of Electronics and Communication Engineering, S V University College of Engineering, Tirupati, Andhra Pradesh, India.

DOIs:10.2015/IJIRMF/202608019     |     Paper ID: IJIRMF202608019


Abstract
Keywords
Cite this Article/Paper as
References

This study presents a comparative evaluation of a conventional convolutional neural network (CNN) and a DenseNet201-based deep convolutional model for single-channel speech enhancement in hearing-assistive applications. The approach uses short-time Fourier transform (STFT) analysis to derive a log-power-spectrum (LPS) representation from noisy speech, with corresponding clean-speech LPS features used as supervised targets. The predicted enhanced magnitude spectrum is reconstructed using the phase of the noisy signal, followed by inverse STFT and overlap-add processing. Performance is evaluated using segmental SNR, SNR improvement, SI-SDR, STOI, PESQ and log-spectral distance (LSD). In the reported experiments, DenseNet201 achieved higher SNR, SI-SDR, SNR improvement, STOI and PESQ than the conventional CNN across the two evaluated speech signals. However, DenseNet201 produced higher LSD, so improvement was not uniform across all spectral measures. The findings support dense feature reuse as a useful architectural strategy for LPS-based enhancement, while highlighting the need for larger held-out test sets, complete reproducibility details and computational profiling before real-time or clinical claims are made.

Speech Enhancement, Hearing Aid, Convolutional Neural Network, DenseNet201, Log Power Spectrum, Noise Reduction, Speech Quality, Speech Intelligibility, PESQ, STOI, SI-SDR.

Dr. S. Swarnalatha, D. Pooja (2026); CNN and DenseNet201-Based Speech Enhancement for Hearing-Assistive Applications, International Journal for Innovative Research in Multidisciplinary Field, ISSN(O): 2455-0620, Vol-12, Issue-8, Available on –   https://www.ijirmf.com/

  1. Bhat, G. S., Shankar, N., Reddy, C. K. A., and Panahi, I. M. S., “A real-time convolutional neural network based speech enhancement for hearing impaired listeners using smartphone,” IEEE Access, vol. 7, pp. 78421–78433, 2019. doi:10.1109/ACCESS.2019.2922370.
  2. Sun, Z., Li, Y., Jiang, H., Chen, F., Xie, X., and Wang, Z., “A supervised speech enhancement method for smartphone-based binaural hearing aids,” IEEE Transactions on Biomedical Circuits and Systems, vol. 14, no. 5, pp. 951–960, 2020. doi:10.1109/TBCAS.2020.2988121.
  3. Strake, M., Defraene, B., Fluyt, K., Tirry, W., and Fingscheidt, T., “Speech enhancement by LSTM-based noise suppression followed by CNN-based speech restoration,” EURASIP Journal on Advances in Signal Processing, vol. 2020, art. 49, 2020. doi:10.1186/s13634-020-00707-1.
  4. Soleymanpour, R., Soleymanpour, M., Brammer, A. J., Johnson, M. T., and Kim, I., “Speech enhancement algorithm based on a convolutional neural network reconstruction of the temporal envelope of speech in noisy environments,” IEEE Access, vol. 11, pp. 5328–5336, 2023. doi:10.1109/ACCESS.2023.3236242.
  5. Huang, G., Liu, Z., Van Der Maaten, L., and Weinberger, K. Q., “Densely connected convolutional networks,” Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp. 4700–4708, 2017.
Post Views: 10

Download Full Paper

Download PDF No. of Downloads:6 | No. of Views: 9