An Unusual Voice Detector Based on Nursing at Home
Chao Chun Wang
Ling Hwei Chen
|關鍵字:||音訊分類;看護系統;Audio classification;Nursing system|
|摘要:||由於工業社會的快速發展和高齡化的社會結構產生越來越多的獨居老人，使得獨居老人缺乏妥善照顧。於是有發展居家看護系統的想法。就是在獨居老人的週遭裝設偵測儀器，即時傳送老人的生理狀況，進而掌握老人的健康情形。目前的看護系統大多是利用視訊或電子相關儀器來偵測被看護者之狀態，而本篇論文主要是從被看護者發出的聲音來偵測是否被看護者的健康情況有疑慮。首先我們定義四種危急聲音，分別為：劇烈咳嗽聲、呻吟聲、喘息聲和求救聲。當被看護者發出以上之危急聲音時，我們判定他的健康情況有疑慮，需有人實際去關注。為偵測上述的四種危急聲音，我們從聲波圖上擷取區域內聲音區段的數目 ( Number of the Segment Parts ) 、波形持續時間 ( Duration of Waveform ) 、平均音量 ( Mean of Volume ) 、正規化零越率 ( Zero Crossing Rate ) 和聲音區段彼此之間的相關性 ( Correlation ) 五種具區別率的特徵為我們偵測之依據。實驗結果顯示，對四種危急聲音的偵測率高達94%~97%，且假警報只有0.08%。|
Because fast development of the industrial society and structure of aging society, there are more and more solitary people. Thus, developing a nursing system at home is necessary. That is some instruments of detection are installed in the solitary people’s home, the physiological states of the old man are conveyed immediately, and the health situation of the old man is under control. Most nursing systems use video information or electronic instrument to detect the healthy state of the person. In this method, we hope to know whether the health condition of the person nursed has a doubt by detecting from the voice that the person nursed emitted. We define four kinds of unusual voices at first, including cough, groan, wheeze and cry for help. When the person nursed sends out the above four kinds of unusual voices, we judge that his health condition have a doubt, and need someone to pay attention actually. In order to detect four kinds of unusual voices, we extract five features on audio waveform, including the number of segmented parts, duration of waveform, mean of volume, zero crossing rate and correlation. Experimental results show that the detection rate can be up to 94%~97% for these four kinds of unusual voices. In false alarm, there are only 0.08% of wrong rates.
|Appears in Collections:||Thesis|
Files in This Item: