Research And Implementation Of Audio Quality Evaluation And Speech Recognition Preprocessing Technology

In recent years, with the popularization of mobile terminal equipment, people's voice communication between more and more frequent, so the quality of audio has gradually taken up. Due to the uneven quality of audio produced in different environments, poor quality audio can have a large impact on people's auditory perception, so the audio quality is evaluated accurately and effectively enhanced for poor audio Getting more and more important. At home and abroad for audio quality evaluation and voice preprocessing technology to do a lot of research work, there are some fruitful research results are presented.However, there are still few feasible schemes for audio quality evaluation,and there is no standard feasible scheme, and the related research results are relatively insufficient. In the audio quality enhancement, the researchers at home and abroad put forward some solutions, but these programs are still relatively large limitations, can only be applied in some specific scenarios, less versatile.In view of the above problems, the following research work is done in this paper: First, the purpose and significance of the audio quality evaluation are introduced. Secondly, the existing audio quality evaluation scheme is studied deeply, and the principles and advantages of the existing schemes are analyzed. Finally, the paper puts forward a set of effective evaluation scheme based on the existing scheme and introduces the related calculation method.In the research of speech recognition preprocessing technology, this paper firstly studies the well-known methods in speech enhancement technology, and has a detailed understanding of the principle and advantages and disadvantages of these methods. Then, based on Speex open source Improved Speech Enhancement Method for Short - Term Spectral Estimation. This method can eliminate the background noise effectively and avoid the appearance of the level noise. For different audio application scenarios, the system achieves an ideal balance between the objective recognition rate and the human subjective listening test, and the system can be freely set Noise reduction, the amplitude of human voice enhancement and other parameters to improve the flexibility of the audio pre-processing system, compared to the traditional method of speech noise reduction, speech recognition in the accuracy rate of about 13%.