Design Of DSP-based Digital Audio Processing System

DSP technology is more and more widely used in the field of audio processing.At present, a number of voice processing systems have been used in speech analysismodule, collected on-site voice and spectrum analysis. Real-time voice processingsystem, power consumption, size, as well as the voice signal fidelity are the keyfactors that affect system performance. Therefore, the audio signal analyzer design isnecessary.As technology advances, the digital technology has been deep into people’s lives,digital technology has an unparalleled advantage of many simulation technology. Thetraditional audio technology can not meet the requirements of the people, digitizedaudio processing method for the audio technology trends. The rapid development ofdigital technology, in many cases makes digital audio technology to gradually replacethe analog audio technology, digital audio processing using digital filtering algorithmtransform processing of the collected signal.Using digital technology, it will surely involves a lot of number crunching, andthe emergence and development of digital computing processor (DSP) is to meet thisrequirement. Many of the special structure of the DSP chip can quickly achieve avariety of digital signal processing algorithms. Digital signal processing technologyhas become a mainstream subject, more and more extensive. Audio widely exist inreal life, it is important for audio processing. The audio signal into a digital signal,and how to deal with these signals, do a simple expounded in this paper.The paper first introduces the voice analysis system works based onTMS320C5402DSP chip, given the overall design of the program and the blockdiagram, and then gives the hardware design of the system; then introduced basedTMS320C5402DSP chip voice recording system software design. Wavelet transformand multiresolution analysis as the theoretical basis of speech endpoint detection inwavelet coefficient variance algorithm and sub-band average energy algorithmanalysis and research, and take advantage of the differences in the frequency domain of speech and noise, these two algorithms optimized and applied to the endpointdetection system, effectively improve the wavelet coefficient variance algorithm istime-consuming, the disadvantage of poor real-time, and to overcome the sub-bandaverage energy algorithm only white Gaussian noise limitations detect the effect ofgood voice endpoint detection The practicality of the system.Throughout the design process, we used TLV320AIC23DSP chip the core audioplayback interface device, the hardware design the combined TMS320C5402DSPchip voice data storage FLASH memory. The software part of a modular designmethod, using the C language. The voice recording designed to achieve the followingfunctions:1. Speech acquisition, the hardware design of the DSP chip, complete voicehardware, you can get a collection of voice information.2. Voice storage, voice storage, the required sampling signal for sampling theanalog voice input signal, the sampling rate of8K/s. Completed by the built-in A/Dconverter chip.3. Voice filtering in the design, the first to use MATLAB simulation produces atime-domain waveform signal observed before filtering and frequency domainwaveforms. By MATLAB simulation, the use of the filter parameters,re-programming a DSP processingSpectrum analysis, spectral analysis by MATLAB simulation and Cool Editsoftware, and on this basis, the use of average weight adjustment method, the strongalignment of the weighting method, self-alignment of the weighting method analysisof simulation results.