Investigation of an impact of room acoustics on performance of ideal binary mask

Speech signals separation or extraction from noise is becoming more and more important in variety applications such as mobile phones, conference equipments, radars and others. That is the reason why this paper is focused on investigating how the environment in which speech is generated can affect performance of separation technique. For this analysis, it has been chosen technique based on ideal binary mask (IdBM) and room acoustics has been simulated by the room impulse response generator. The present study replicates and extends the findings which have been already presented but mainly shows impact of room acoustics on performance of IdBM. Several different tests have been performed for varying values of reverberation time, distance and type of microphone. Although subjective evaluation measures are more accurate, the objective measures have been used in this paper because they are less time consuming, directly computed and can be used repeatedly. Only objective measures with good correlation with subjective measures have been used for higher accuracy of results such as PESQ or STOI.