利用人臉識別探討懷舊中文歌曲對於長期照顧機構住民情緒的影響

Using Face Recognition to Explore the Emotional Impact of Nostalgic Chinese Song on the Residents in Long-term Care Facility

背景與目的：懷舊音樂可以舒緩壓力與減低焦慮，且與正向情緒連結有關。長期照顧機構住民因孤寂感而造成情緒低落，懷舊音樂能幫助舒緩情緒。本研究目的是利用人臉識別科技，探討懷舊中文歌曲於機構住民情緒之關係。方法：本研究設計為立意取樣，透過機器深度樣本學習，擷取台北市某長期照顧中心15位個案聆聽120秒中文老歌後的情緒變化。面部表情資訊共收集25,200筆資料，利用成對樣本t檢定、無母數魏克生符號等級檢定、卡方檢定等統計方法進行分析。結果：成對樣本t檢定結果顯示，個案聆聽中文老歌變化，高興(Happy)情緒有顯著增加(前測平均值=0.15，標準差=0.12；後測平均值=0.27，標準差=0.15； p < 0.05)；驚訝(Surprised)情緒則顯著降低(前測平均值=0.04，標準差=0.03；後測平均值=0.02，標準差=0.01；p < 0.05)。以無母數魏克生符號檢定，發現受試者在聆聽歌曲後的情緒表現與成本樣本t檢定一致。卡方檢定顯示所有受測者聆聽歌曲的每秒所測得情緒變化人次比例均有顯著的差異，尤其以高興(Happy)人次增加比例最高，從302人次（17.09%）增加到577人次(32.36%)。結論：人臉辨識科技驗證懷舊中文老歌能幫助長期照顧住民正面情緒變化，建議長照機構能利用懷舊音樂輔療達到舒緩住民情緒的目的。

 

Objective: Music can relieve stress and reduce anxiety and is related to positive emo¬tions. Residents in a long-term care facility feel depressed because of loneliness, so nostalgic music can help soothe their mood. This study aimed to explore the relationship between nostalgic Chinese songs and the emotions of residents using facial recognition technology. Methods: This study used purposive sampling and the machine learning to train the samples. The facial recognition catches the changes among 7 emotions in 120s by the pretest without music and with the intervention of nostalgic music in an assisted-living facility in Taipei. A total of 25,200 pieces of data were collected in 15 cases and analyzed by a paired sample t-test, Wilcoxon signed-rank test, and Chi-square test. Results: The happy emotion significantly in¬creased when listening to nostalgic Chinese songs (pre-test Mean = 0.15, SD = 0.12; post-test Mean = 0.27, SD = 0.15; p <.05) but the surprised emotion was significantly reduced (pre-test Mean = 0.04, SD = 0.03; post-test Mean =.02, SD = 0.01; p <.05) by the paired sample t-test. When compared with the Wilcoxon signed-rank test, the emotional performance after listen¬ing was consistent with the paired sample t-test. When all the residents listened to the music, there were significant differences by the Chi-square test in the measure of emotional ratio per person per second. In particular, the increase in the happy emotion was the highest, from 302 (17.09%) to 577 (32.36%). Conclusion: Facial recognition proved that nostalgic old Chinese songs could help promote positive emotional changes in the residents of a long-term care fa¬cility. It is recommended to use nostalgic music therapy to alleviate residents’ emotional distress.