Objective Electrolarynx (EL) is the most common device to provide a voice for laryngectomees. However, the EL phonation has a serious drawback: the radiated noise. To improve the speech quality by eliminating the radiated noise becomes a goal for researchers. Spectral subtraction method is the most common way to eliminate the radiated noise of EL whose pitch is constant. However, for the pitch adjustable EL speech, the effectivity of this method is not known. Thus, this paper conducts a study on the noise reduction effects of different spectral subtraction methods for pitch adjustable EL speech. Methods The classical spectral subtraction, the multi-band spectral subtraction, the perceptual weighting spectral subtraction and the weighting function spectral subtraction were firstly introduced, respectively. Then, a male native speaker of Mandarin Chinese who had been trained to be familiar with using pitch-controlled EL was instructed to read twenty daily mandarin sentences in a soundproof room, which was recorded as the raw data of the EL speech. Furthermore, the methods above were used to remove the radiated noise of EL speech, and subjective and objective methods were used to evaluate and compare the enhanced speech. Results Objective results showed that the enhanced EL speech by classical spectral subtraction still remained noise obviously though a part of radiated noise was reduced. The other three methods including multi-band spectral subtraction, the perceptual weighting spectral subtraction and the weighting function spectral subtraction reduced the noise effectively, especially in speech-pause section. Nevertheless the enhanced EL speeches of the former two methods still had a little noise in speech section. Meanwhile, certain high frequency speech components of the normal speech might be eliminated by the weighting function spectral subtraction. Subjective results indicated that compared with the original EL speech, the acceptability of the enhanced speeches by different spectral subtraction methods increased and the weighting function spectral subtraction method obtained the highest score. Yet the intelligibility of the enhanced speeches had little change. Conclusions According to the selected EL speech whose pitch is adjustable, the enhancement methods based on spectral subtraction can effectively reduce the radiated noise of EL speech and improve the acceptability and speech quality, among which the weighting function spectral subtraction method obtains the best performance. However, these methods have little influence on speech intelligibility.