A method (100) and apparatus (200) are disclosed for transcribing a humming signal into a sequence of musical notes. The method begins by grouping (305) the signal into frames of data samples. Each frame is then processed to derive (320) a frequency distribution for each frames. The frequency distributions are processed to derive (410) a Harmonic Product Energy (HPE) distribution over the frames. The HPE distribution is then segmented (115, 120) to obtain boundaries of musical notes. The frequency distributions of the frames are also processed to derive (412) a fundamental frequency distribution. A pitch for each note is determined (125) from the fundamental frequency distribution.