Research And Application Of Audio Sample Rate Conversion Algorithm

The analog real-world and digital signal processing tools makes sampling be thekey way to get a digital signal from the analog source. With the rapid development ofinformation technology, according to the characteristics of the signal itself and theneeds of processing, many different sampling rates came into being. In audio world,there are a variety of sampling rates, such as48kHz for digital audio tape (DAT)，44.1kHz used in CDs, and32kHz for digital audio broadcasting. In practice, we oftenencounter the issue of sample rate conversion. Sample rate conversion is a process ofincreasing or decreasing samples of the unit time. For the last20years, the samplerate conversion theory and its implementation has become an important branch ofmodern digital signal processing with broad application prospects. State-of-the-artsample rate conversion algorithms effectively solve the problem of some kinds ofaudio sample rate conversion in terms of performance, processing speed and memory,while they are not so efficient for other kinds. Thus, to construct a simple but efficientsystem to accomplish all kinds of audio sample rate conversion is an issue worthy ofstudy.Therefore, this article studies traditional sample rate conversion algorithm andFarrow-based FIR fractional delay filter algorithm. Traditional sample rate conversionalgorithm applies to most kinds of audio sample rate conversion, and farrow-basedFIR FD filter applies to audio sample rate conversions containing44.1kHz.Farrow-based FIR FD filter has two deficiencies, narrow passband and imageattenuation limited, thus, an improved structure that combines an integer upsamplewith Farrow-based FD filter is adopted.Then, an audio sample rate conversion system which supports the conversionamong12kinds of sample rate conversion is designed. The whole sample rate isdivided into two stages. First stage implements polyphase decimation or interpolation,while second stage uses Farrow-based FIR FD filter. The overall sample rateconversion system is simulated and verified with Matlab.Finally, write codes and implements the whole system in VC, transplant theprogram into an embedded platform MIPS324KEc and optimize it in terms of speedand memory. After the comparison and analysis of the experimental results, it comes to the conclusion that the system proposed is capable of implementing the real-timeand efficient conversions among12kinds of audio sample rates, and the results issatisfactory.