Research And DSP Implementation Of Speech Enhancement Algorithms Based On Microphone Array

Speech enhancement is to restrain the noise interference involved in the received speech signals as efficiently as possible. Speech enhancement using a single microphone cannot reduce noise more desirably especially in the environment with strong reverberations because it only provides time/frequency information. In contrast, speech enhancement methods with microphone arrays also make full use of information in the space domains to get better results of speech enhancement.Time Delay Estimation(TDE) is a key part in most speech enhancement methods based on microphone array and their accuracy directly affects the results of speech enhancement. CSP and its improved method GCSP can help to obtain good TDE result, but their performance is restrained by the length of the speech frame and when more microphones are used the software will get more complicated. To solve this problem, the correlation function based on the coherence function is provided in this paper, which can not only give out more exact TDE result in the case of strong reverberations and low Signal-to-Noise Ratio (SNR), but also reduce the complexity of correlation function when the maximal value of time delay is established.For the real-time application software system, the designer should focus on the cost of hardware resource and its restriction to the software complexity. In this paper, ADSP-BF533 is chosen to be the central processor and we use Visual DSP++ to develop the software system and its real-time simulation.In this paper, the study of basic principles on speech enhancement with microphones array provides a positive effect on the following further research and comparison among kinds of speech enhancement and TDE methods. In addition, a software application scheme for the system of speech enhancement with microphones array using GSC algorithm based on ADSP-BF533 is provided, developed and simulated.