Abstract: Because of the low SNR of the whispers，it is difficult to obtain both high accurate rates and good robustness with some traditional detection algorithms in whispered environment.An algorithm of whispered speech endpoint detection is presented which is based on Instantaneous Energy Frequency Value（IEFV） of Hilbert-Huang Transform（HHT）.This paper applies HHT to separate instantaneous amplitude and instantaneous frequency from whispers，and extracts IEFV，which is a temporal-amplitude-frequency character.Because IEFV can distinguish whispers form noise effectively，it is used as the feature for endpoint detection.The accurate rates of both initial and final of this algorithm are higher than the Zero-Energy-Product method，the Spectral Entropy method and the Fitting Characteristic method in the test with 700 samples at 2～10 dB SNR.As shown in the experiments，this algorithm can detect whispered speech endpoint accurately in various non-stable noisy backgrounds.