基于AI的噪聲環境下遠距離高清音頻采集研究

摘" 要： 為提升遠距離采集音頻信號的強度，深度濾除音頻信號噪聲以提取有用音頻部分，提出一種基于AI的噪聲環境下遠距離高清音頻采集方法。構建遠距離高清音頻采集結構，分別通過模擬增益和數字增益技術進行音頻信號增益處理，以提升音頻信號強度。依據短時傅里葉變換提取音頻增益信號頻域特征，輸入到長短期記憶網絡中，實現音頻信號深度噪聲去除，得到高清音頻頻域信息；再通過短時傅里葉逆變換處理該信號，實現音頻信號重構，最終達到噪聲環境下遠距離高清音頻采集的目的。實驗驗證結果表明：依據音頻信號增益能夠有效提升采集音頻信號的強度，并避免信號受距離、噪聲影響而逐漸衰減，繼而有效濾除音頻信號噪聲數據，提取其中有用的音頻信號，確保音頻信號高清度；且最終采集音頻信號信噪比均高于18 dB，可懂度均高于97%，有效驗證了所提方法的有效性和準確性。

關鍵詞： 高清音頻采集； AI； 噪聲環境； 信號強度； 遠距離； 長短期記憶網絡； 短時傅里葉變換

中圖分類號： TN912.3?34； TP399" " " " " " " " 文獻標識碼： A" " " " " " " " " " " "文章編號： 1004?373X（2025）04?0130?05

Research on AI?based remote HD audio acquisition in noisy environment

HUANG Lina

（Jiangxi University of Science and Technology， Ganzhou 341000， China）

Abstract： In order to enhance the strength of long?distance audio signal acquisition and deeply filter out audio signal noise to extract useful audio parts， a method of AI?based remote high?definition （HD） audio acquisition in noisy environments is proposed. The remote HD audio acquisition structure is constructed， and the analog gain and digital gain techniques are used to conduct the audio signal gain processing， so as to improve the strength of audio signal. The frequency domain features of audio gain signals are extracted based on short?time Fourier transform， which are inputted into long short?term memory network to realize the deep noise removal of audio signals and obtain HD audio frequency domain information. The signal is processed by means of short?time Fourier inverse transform， to realize the audio signal reconstruction， ultimately achieving the goal of remote HD audio acquisition in noisy environments. The experimental verification results show that the gain of the audio signal can effectively enhance the strength of the collected audio signal， and avoid the gradual attenuation of the signal due to distance and noise， effectively filtering out the noise data of the audio signal， extracting useful music signals， and ensuring HD of the audio signal. The final collected audio signal signal?to?noise ratio was higher than 18 dB， and the comprehensibility was higher than 97%， effectively verifying the effectiveness and accuracy of the proposed method.

Keywords： high?definition audio acquisition; AI; noisy environment; signal strength; remote; long short?term memory network; short?time Fourier transform

0" 引" 言

音頻采集在音頻技術中具有重要作用，能夠確保信息的準確傳達，便于教學、會議等后期分析和回顧，提升音頻、娛樂等用戶體驗[1]。實際的音頻采集環境可能較為復雜，但實現噪聲環境下遠距離高清音頻采集能夠捕捉較為清晰的音頻信號[2]，有助于提升音頻信息完整性，為后期的分析和處理提供良好的依據。……