In　this　paper,　we　propose　an　adaptive　jointly　parameter　quantization　scheme　for　sinusoidal　matching　pursuit　(MP)　audio　coding,　where　the　sinusoidal　parameters　(amplitude,　frequency,　phase)　are　quantized　jointly　based　on　the　real　time　statistical　knowledge　of　parameters.　The　optimal　quantizers　which　minimize　the　perceptual　distortion　under　a　given　entropy　constraint,　are　derived　analytically　by　high-rate　theory.　In　comparison　tests,　the　significant　improvements　achieved　with　the　proposed　scheme　are　shown.　©　2010　IEEE.