Ambient noise such as instrument noise and human voices often disturbs the hearing and/or measurement of lung sounds. Conventional frequency-domain filtering is usually ineffective. Noise is transmitted to the microphone that measures lung sounds through the chest wall around it, and it may be feasible to cancel out the noise by identifying this transfer function. The function, however, may vary with respect to the subject and measuring site, and therefore it should be modified dynamically. We apply an adaptive filtering technique to solve this problem. A workstation-based off-line adaptive noise canceller is developed to assess its performance in detail. Filter coefficients are controlled by a least-mean-square algorithm. Results show that the ambient noise is reduced by about 30 dB in a convergence time of several seconds. A real-time adaptive noise canceller is subsequently implemented by incorporating a digital signal processor, and a prototype electronic stethoscope is realised with high immunity to ambient noise. In a clinical application experiment in which the noise-contaminated lung souds are observed during an airway sensitivity test, satisfactory results are obtained. It is proved that the proposed method and device are effective for hearing and/or measuring lung sounds in noisy environments.