Speech　Emotion　Recognition　(SER)　has　become　an　indispensable　part　of　human-computer　interaction.　In　order　to　obtain　a　higher　recognition　rate,　in　this　paper,　a　joint　feature　based　on　Philips　fingerprint　and　spectral　entropy　is　used,　and　it　is　combined　with　some　underlying　features　to　perform　speech　emotion　recognition　on　the　four　emotions　of　angry,　neutral,　happy,　and　sad.　Firstly,　the　Mel　Frequency　Cepstral　Coefficients　(MFCC),　Linear　Prediction　Coefficient　(LPC),　logarithmic　amplitude-frequency　characteristics,　Philips　fingerprints　and　spectral　entropy　are　extracted　for　speech　dataset.　Secondly,　Principal　Component　Analysis　(PCA)　is　used　to　reduce　the　dimension　of　the　extracted　feature.　Thirdly,　SVM　is　used　to　train　and　classify　the　features　after　dimension　reduction.　The　experimental　results　show　that　by　comparing　the　recognition　results　with　other　feature　combinations,　the　method　proposed　in　this　paper　has　good　recognition　results.　©　2022　ACM.