Wide　area　monitoring　systems　(WAMS)　provide　effective　support　for　online　identification　of　low-frequency　oscillations　in　power　systems.　The　WAMS　signal　is　sensitive　to　the　surrounding　environment,　and　it　contains　Gaussian　white　noise.　The　Gaussian　white　noise　will　produce　Gaussian　coloured　noise　through　the　filter,　which　will　bring　some　errors　to　the　mode　identification.　In　order　to　solve　the　above　problem,　this　study　proposes　a　method　to　identify　the　low-frequency　oscillation　modes　of　a　single-channel　measurement　of　the　power　system　based　on　the　combination　of　the　fourth-order　mixed　mean　cumulant　and　the　improved　TLS-ESPRIT　(total　least　square-estimation　parameter　space　rotation　invariant　technique).　The　actual　signal　is　replaced　by　the　processed　signal　using　the　fourth-order　mixed　mean　cumulant　mathematical　statistical　method,　thereby　the　Gaussian　coloured　noise　is　suppressed　effectively;　then,　TLS-ESPRIT　algorithm　is　used　to　identify　the　signal,　which　is　used　to　identify　the　low-frequency　oscillation　of　each　mode.　A　novel　relative　change　rate　of　singular　value　was　also　introduced　in　order　to　avoid　the　artificially　threshold-setting　error.　The　simulation　results　show　that　the　proposed　method　has　better　anti-noise　performance　and　higher　accuracy　of　fitting　than　other　methods　reported　in　literature,　besides　being　easy　to　be　implemented　and　suitable　for　performing　on-line　identification.