Research And Implementation Of Speech Enhancement On FPGA

The primary purpose of speech enhancement is to purify it from noise. Because it is very complicated to grasp the random noise's cause or characters, it is impossible to eliminate the random noise completely. Actually, goals of speech enhancement are to improve the quality of speech, to eliminate the background noise so that every hearer is willing to hear and wouldn't feel weary, and to upgrade the quality of speech and understandable level.In this paper, firstly, several speech enhancement algorithms will be introduced, such as the least-mean-square(LMS), normalized-LMS, sign-error-LMS and recursive least square(RLS) adaptive filter algorithms. These filters minimize the difference between the output signal and the desired signal by altering their filter coefficients. Over time, the adaptive filter's output signal more closely approximates the signal you want to reproduce not only the fundamentals of these methods were stated, but also the ways they can be put into practice, some improvement was made, experiment results were compared ,and the connections and differences were also revealed in this essay.On the basis of research of adaptive filter theory, LMS adaptive filter algorithms especially studied. For the different noise, the astringency,convergence speed And stability of the speech enhancement algorithms weight coefficient is especially analyzed .From the simulation result, RLS adaptive filter algorithm have a good performance to deal with the stationary noise and the non-stationary noise .but to the LMS adaptive filter such as the same speech noise, speech enhancement effect isn't perfect in the non-stationary noise.The principle of FPGA are introduced, especially about ALTERA CYCLONE2. The author designed a speech enhancement system which is successful both in hardware and software design and this system can restrain noise from noisy speech effectively.