Microphone Array Speech Processing And OMAP-L137 Platform Implementation

In the acoustic system such as hands-free telephone and teleconference,besides the interference of environmental noise,the coupling between microphone and speaker also introduces new acoustic echo interference problem.In addition,the current voice terminal equipment is gradually developing towards miniaturization and low power consumption,so it is of great significance to study the speech enhancement and echo cancellation technologies which are easy to be deployed in specific hardware platforms.Based on the existing hardware platforms,this thesis studies the beamforming and acoustic echo cancellation algorithm based on microphone array.The main works are as follows:(1)The defects of traditional beamforming methods based on additive array are summarized.Based on the characteristics of speech signals,a frequency-invariant beamforming method of circular array is proposed,which can not only solve the space limitation problem of small terminal speech devices,but also ensure high directional gain independent of frequency.In order to effectively improve the influence of room reverberation on speech quality,complex cepstrum theory is introduced to improve the robustness of the algorithm.Finally,the frequency invariant characteristics of the proposed model are verified by experiments.The results show that the proposed method can extract clear desired speech signals in noisy environment based on OMAP-L137 hardware platform,and has superior performance in speech enhancement.(2)Aiming at the limitations of traditional acoustic echo cancellation methods,an improved microphone array echo cancellation method is proposed.The proposed method fully considers the problems of adaptive algorithm in acoustic echo path estimation,effectively utilizes the spatial filtering characteristics of array,and gets rid of the dependence of traditional acoustic echo cancellation methods on double-talk detection module.At the same time,the traditional Improved Proportionate Normalized Least Mean Square(IPNLMS)algorithm is improved to further improve the convergence speed and reduce the computational complexity.Experiments show that the improved algorithm based on OMAP-L137 hardware platform has good performance of noise reduction and echo cancellation for noisy speech signals in the presence of acoustic echo.(3)The realization of speech signal processing technology in Digital Signal Processor is studied.Considering the miniaturization and low power consumption of the system,a fully functional voice hands-free system is designed and implemented based on OMAP-L137 hardware platform,which makes the related technologies studied in this thesis get specific application.The System uses the Voice over IP(Vo IP)architecture to realize the two-terminal hands-free call,and develops a friendly and interactive user interface based on the Vue2 framework.The test data show that the designed system can ensure the clarity and intelligibility of speech quality in the actual hands-free call scene.