Research on Adaptive Coding and Denoising System Based on Sound Quality and Audio Effect Balance Optimization

Yechi Zhang

doi:10.54097/tr75p867

Authors

Yechi Zhang

DOI:

https://doi.org/10.54097/tr75p867

Keywords:

Entropy Weight Method TOPSIS, Multidimensional Audio Evaluation Model, MDCN, LSTM Architecture.

Abstract

This article proposes a comprehensive evaluation index, cost-effectiveness analysis model, adaptive encoding scheme, and improved denoising strategy for optimizing audio storage and noise removal. To improve audio quality and storage efficiency, and to quantify the balance between storage efficiency and sound quality fidelity for different audio formats, this article establishes a sound storage balance quantization model. By constructing a comprehensive evaluation quantization function and combining multidimensional factors to evaluate the balance between storage efficiency and sound quality fidelity. Based on a three-level indicator system (basic technology, performance evaluation, and comprehensive evaluation), entropy weight method is used to calculate indicator weights, and TOPSIS method is applied for comprehensive scoring. The results indicate that AAC format has the best sound storage balance performance, while WAV and FLAC have superior sound quality but lower storage efficiency, MP3 focuses more on storage efficiency in low bandwidth environments. At the same time, this article establishes a multidimensional audio evaluation model that quantifies sound quality using factors such as mean square error (MSE), while quantifying the size of audio files through factors such as audio duration, sampling rate, bit depth, number of channels, and compression ratio. Based on relevant features, the sound quality performance of speech and music content was evaluated, and the comprehensive and relative effects of sampling rate, bit depth, and compression algorithm on sound quality were analyzed. Based on this, an audio performance balance index was designed to quantify the cost-effectiveness between sound quality and file size. The sorting results show that music-44100Hz-MP3_320kpp.mp3 and speech-44100Hz-MP3_320kpp.mp3 have the highest cost-effectiveness, and recommend the best parameter configurations for voice and music audio.

References

[1]Chen Qingyun, Gai Junmei Research on the Selection and Conversion Path of Audio File Archiving Format [J]. Lantai World, (06):50-53, 2022.

[2]Hang, T., Gu, S., Li, C., Bao, J., Chen, D., Hu, H., Geng, X., Guo, B. (2023). Efficient Diffusion Training via Min-SNR Weighting Strategy. CVF, 7441: 1-8, 2023.

[3]Geng Lele A Study on the Equity of Per Capita Expenditure on Basic Education in China: Calculation Based on Gini Coefficient and Theil Index [J]. Journal of East China Normal University (Education Science Edition), 40 (01): 60-73, 2022.

[4]Wang Lu Research on Anomaly Detection Algorithm Based on Multi Model Integration [D]. Shandong University, 2023.

[5]Shi Guigang, Liu Wanduan, Zuo Jingyao Research and Comparative Analysis of Index Weighting Techniques Based on Entropy Weight Method [J]. Journal of Jiamusi University (Natural Science Edition), 2024, 42 (11): 168-171, 2024.

[6]Yang Liang, Hou Wei Error analysis of multi-source 3D geological modeling data based on entropy weight method [A] Proceedings of the 12th National Symposium on Mathematical Geology and Geological Information [C]: 391, 2013.

[7]Sun Dajun, Ma Chao, Mei Jidan, etc Vector array deconvolution beamforming method based on non negative least squares [J]. Journal of Harbin Engineering University, 40 (07): 1217-1223, 2019.

[8]Lin, T.-Y., Goyal, P., Girshick, R., He, K., Dollár, P. (2017). Focal Loss for Dense Object Detection. IEEE, pp. 1-9, 2017.

[9]Li, F., Guo, S. (2024). Multi-Scale Dilated Convolution Network for Long-Term Time Se- ries Forecasting. ICLR, pp. 1-9, 2024.

[10]Kang Ning, Mo Luyao, Jing Ke (2024). Semi parametric smooth transformation quantile autoregressive model and its application Systems Science and Mathematics: 844, 2024.