To　evaluate　underwater　sound,　a　simple　non-invasive　optical　technique　based　on　self　-interference　to　detect　lowfrequency　underwater　acoustic　signals　was　demonstrated.　Clear　self　-interference　fringes　of　a　laser　beam　reflected　from　the　surface　capillary　wave　transformed　from　a　low　-frequency　underwater　acoustic　signal　by　a　cylinder　were　observed.　This　study　was　also　aimed　at　developing　an　artificial　neural　network　(ANN)　model　with　input　from　an　optical　pattern.　The　relationship　between　the　optical　pattern　and　the　frequency　of　underwater　sound　was　established,　and　an　optimal　combination　of　hyper　-parameters　was　obtained.　By　analyzing　the　fringe　region　and　the　fringe　interval,　the　frequency　of　the　underwater　acoustic　signal　and　its　relative　amplitude　were　measured.　A　model　based　on　physical　optics　modified　the　fringe　distribution　function,　and　the　theoretical　fit　with　the　modified　function　was　in　good　agreement　with　experimental　observation.　The　BP　-ANN　model　exhibited　good　performance　in　establishing　the　relationship　between　the　and　underwater　sound.