A Study of the Sound Quality Analysis of Singer Based on a DSP

• Main Authors: Yu-Chi Tsai, 蔡煜基
• Other Authors: Wei-Chih Hsu
• Format: Others
• Language: zh-TW
• Online Access: http://ndltd.ncl.edu.tw/handle/84086216412830819849

碩士 === 國立高雄第一科技大學 === 電腦與通訊工程所 === 92 === Nowadays, the karaoke equipments are very popular with people for entertainment purposes. Some of them have the score function to attract more people to participate in this activity. However, this function makes the score less accurate. Thus, the louder the singer sings the higher the score is. Yet, according to the explanation of the analysis of domain, it’s unable to determine the quality of the sound accurately through the amplitude, which is produced by the time domain singing sound sample. Therefore, the main purpose of this research is not only to improve the above weaknesses but also to analyze the characteristics of the stunning sound quality. In his book “In Search of Beautiful in Music”, Carl E. Seashore, the professor of University of Iowa, thoroughly analyzed the sound quality of the vocalist. He explained the timbre aspect, which measures the distribution and the energy of the fundamental wave and harmonics of different sound. Moreover, he also described the pitch periodic change in the sonance aspect (vibration change). Therefore, this research, based on the above report, implements the system to analyze the sound quality. This research utilizes C54xxDSK board whose core contents of C5402 digital signal processor. C5402 is 16 bits fixed digital signal processor. With pipeline structure, C5402 is capable to operate, in high rate, multiplications and additions in one instruction cycle. The memory-mapping, which is used of C5402’s register, has the quick access characteristic. It provides the Direct Memory Access control (DMA), unify the double buffer memory area of operating mode, and achieve the simultaneous reception and analysis sample. The Butterworth design methods, consisting passband flat nature, designs a simple anti-alias filter. At first, the operation raises the sample rate (oversampling), and then designs a precise FIR of digital low pass filter. Afterward the operation falls the sample rate ( downsampling). By means of these three steps, we are able to eliminate the no-nlinear distortion of analogy filter caused by the non-linear phase. The multiplex sample rate processing method by utilizing the decimation and interpolation of operation technique designs an interpolation FIR filter to which matches the narrow transition band specification. This method effectively reduces the filter step number, makes the system work efficiently, and has saved the memory space. In this research, Algorithms, which transfers the time domain to frequency domain, employs the Fourier Transform. Comparing to Discrete Fourier Transformation, Fast Fourier Transform, the high efficient algorithm, can reduce the time of the multiplications and additions. In the programming optimization process, related to the floating point value using in the fixed point digital processing system, operates in three parts: 1. The data flow analysis 2. The module analysis 3. The Q-formatting shift It makes fixed point processing more accurately. The programming module architecture and simplification as well as memory allocation are properly assigned and multi-analyzes the disposition. Therefore, this research mainly analyze the sound quality by the following aspects: 1. Using the transformation to obtain the distributing of the line spectrum magnitude. 2. Analyzing the singing sound samples and the frequency domain. 3. Computing the fundamental wave and harmonic numbers and the energy. 4. Selecting the maximum frequency of fixed-length frame 5. Observing a set of sound frame frequency change. 6. Calculating standard deviation values. 7. Picking up the sound quality characteristic values The analysis quantification results, by using the series communication port passing on to exterior Display Unit (PC), consists of the complete and practical sound quality analysis system.