In　response　to　the　prevalent　symptoms　of　depression,　the　complexity　of　diagnostic　processes,　and　individuals＇　resistance　to　psychological　testing　in　contemporary　society,　this　study　proposes　an　automated　depression　detection　method　based　on　audio　analysis.　This　method　aims　to　enhance　the　accuracy　of　depression　state　recognition　by　comparing　the　efficiency　of　different　models,　including　machine　learning　and　deep　learning.　Initially,　the　study　collected　audio　sample　data　from　50　individuals,　including　both　healthy　subjects　and　patients　with　depression.　During　the　audio　data　processing　phase,　features　related　to　depression　recognition,　such　as　Mel-frequency　cepstral　coefficients　(MFCC)　and　fundamental　frequency,　were　extracted　from　these　samples.　Subsequently,　a　series　of　recognition　comparison　experiments　were　conducted　based　on　the　extracted　features,　involving　various　algorithmic　models　such　as　Support　Vector　Machine　(SVM),　Random　Forest　(RF),　and　Convolutional　Neural　Network　(CNN).　Experimental　results　demonstrated　that　the　Convolutional　Neural　Network　exhibited　higher　accuracy　in　recognizing　depression　states　from　audio　data,　with　an　average　recognition　accuracy　rate　of　95.82%.　This　finding　indicates　that　deep　learning　models,　especially　Convolutional　Neural　Networks,　have　significant　advantages　in　addressing　such　issues,　providing　robust　technical　support　for　the　future　automatic　detection　of　depression.　©　2024　IEEE.