A new F0 (fundamental frequency) extraction algorithm is proposed, which does not introduce extraction error in principle. The key is to define F0 as the instantaneous frequency of the fundamental component of the signal. This seemingly contradictory definition is made practical by introducing a new ‘fundamentalness’ measure. The measure is defined based on wavelet analysis using an iso-metric Gabor function. A series of evaluation using a database of simultaneously recorded sentence materials with EGG (Electro Glott-Graph) signals was conducted. The performance of the proposed method outperformed conventional methods without any post-processing for a 40Hz to 800Hz F0 search range. For example, over 50% of analysis freme of a 100 sentence female speech databese showed that the agreement is within ±0.3%.