Scholarlink

Human speech compression remains a fundamental challenge in modern communication systems, where bandwidth efficiency must be balanced against intelligibility preservation. This paper presents a novel wavelet-based file format CSA that simultaneously addresses noise suppression and data compression through an integrated Discrete Wavelet Transform (DWT) and Zstd pipeline. The proposed system exploits the inherent sparsity of speech signals in the wavelet domain, achieving a dual objective of noise removal and compression efficiency. The format excels in forensics, voice message storage, and bandwidth-constrained transmission scenarios. Comparative benchmarks derived from tests on a 1000 random audio samples from the Mozilla Common Voice dataset demonstrate superior performance against the most popular encoding formats like MP3, OGG, and AAC in both noisy channel conditions and storage-constrained environments.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Novel File Format for Denoising and Compressing Speech Signals

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Saurav Pal,
  • Ranjushree Pal

摘要

Human speech compression remains a fundamental challenge in modern communication systems, where bandwidth efficiency must be balanced against intelligibility preservation. This paper presents a novel wavelet-based file format CSA that simultaneously addresses noise suppression and data compression through an integrated Discrete Wavelet Transform (DWT) and Zstd pipeline. The proposed system exploits the inherent sparsity of speech signals in the wavelet domain, achieving a dual objective of noise removal and compression efficiency. The format excels in forensics, voice message storage, and bandwidth-constrained transmission scenarios. Comparative benchmarks derived from tests on a 1000 random audio samples from the Mozilla Common Voice dataset demonstrate superior performance against the most popular encoding formats like MP3, OGG, and AAC in both noisy channel conditions and storage-constrained environments.