• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.16 no.7
• /
• pp.1370-1376
• /
• 2012

This paper proposes a gender recognition algorithm which classifies a male or female speaker. In this paper, characteristic vectors for the male and female speaker are analyzed, and recognition experiments for the proposed gender recognition by a neural network are performed using these characteristic vectors for the male and female. Input characteristic vectors of the proposed neural network are 10 LPC (Linear Predictive Coding) cepstrum coefficients, 12 LPC cepstrum coefficients, 12 FFT (Fast Fourier Transform) cepstrum coefficients and 1 RMS (Root Mean Square), and 12 LPC cepstrum coefficients and 8 FFT spectrum. The proposed neural network trained by 20-20-2 network are especially used in this experiment, using 12 LPC cepstrum coefficients and 8 FFT spectrum. From the experiment results, the average recognition rates obtained by the gender recognition algorithm is 99.8% for the male speaker and 96.5% for the female speaker.

• Speech Sciences
• /
• v.10 no.2
• /
• pp.57-64
• /
• 2003

The cepstrum coefficients are the most popular feature for speech recognition or speaker recognition. The cepstrum coefficients are also used for speech synthesis and speech coding but has major drawback of long processing time. In this paper, we proposed a new method that can reduce the processing time of FFT cepstrum analysis. We use the normal ordered inputs for FFT function and the bit-reversed inputs for IFFT function. Therefore we can omit the bit-reversing process and reduce the processing time of FFT ceptrum analysis.

• Journal of IKEEE
• /
• v.16 no.3
• /
• pp.189-196
• /
• 2012

This paper proposes new phase error compensation method of RVDT encoder in the FFT domain. Phase errors are measured with a small combination of compensation resistors and the changes of first order coefficients of FFT for each resistor are obtained. It is found that the coefficient change is inversely proportional to the inserted resistor. The proposed method takes less time and the size of the table is smaller than previous time domain approaches. In addition, the location of the compensation resistor can be found through axis transformation of the coefficients. Finally, the peak-to-peak phase error was improved to 0.57 which is two times better than previous one.

• Journal of the Institute of Electronics Engineers of Korea SD
• /
• v.49 no.3
• /
• pp.8-14
• /
• 2012

In this paper, a low-area FFT structure using $Radix-4^2$ algorithm is proposed. The large point FFT structure consists of cascade connection of the many stages. In implementation of large point FFT using $Radix-4^2$ algorithm, stages which number of different coefficients are only 3 appear in every 2 stages. For example, in the 4096-point FFT, the stages that number of different coefficients are 3 appear in stage 1, 3, and 5 among 6 stages. Multiplication block area of these 3 stages can be reduced using CSD(Canonic Signed Digit) and common sub-expression sharing techniques. Using the proposed structure, the 256-point FFT is implemented with the Verilog-HDL coding and synthesized by $1.971mm^2$ cell area in tsmc $0.18{\mu}m$CMOS library. This result shows 23% cell area reduction compared with the conventional structure.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.15 no.10
• /
• pp.2093-2099
• /
• 2011

Speech signal and car noise signal such as muffler noise are segregated from the one which has both signals mixed using statistical method. To classify speech signal from the other in segregated signals, FFT coefficients were obtained for all segments of a signal where each segment consists of 128 elements of a signal. For several coefficients of FFT corresponding to the low frequencies of a signal, autocovariances are calculated between coefficients of same order of all segments of a signal. Then they were averaged over autocovariances. Linear equation was eatablished for the those autocovariances using the linear regression method for each siganl. The coefficient of the slope of the line gives reference to compare and decide what the speech signal is. It is what this paper proposes. The results show it is very useful.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.46 no.4
• /
• pp.157-164
• /
• 2009

An efficient pipeline MDC Radix-4 FFT structure is proposed in this paper. Every stages in pipeline FFT structure consists of delay' commutator and butterfly. Proposed butterflies in front and rear stages utilize CORDIC and Common Sub-expression Sharing(CSS) techniques, respectively. It is shown that proposed butterfly structure can reduce the number of adders through sharing common patterns of CSD type coefficients. The Verilog-HDL modeling and Synopsys logic synthesis results that the proposed structure show 48.2% cell area reduction in the complex multiplication part and 22.1% cell area reduction in overall 256-point FFT structure comparison with those of the conventional structures. Consequently, the proposed FFT structure can be efficiently used in various OFDM systems.

• The Journal of Korea Institute of Information, Electronics, and Communication Technology
• /
• v.2 no.2
• /
• pp.23-28
• /
• 2009

This study describes an emotion classification method using two different feature parameters of four-channel EEG signals. One of the parameters is linear prediction coefficients based on AR modelling. Another one is cross-correlation coefficients on frequencies of ${\theta}$, ${\alpha}$, ${\beta}$ bands of FFT spectra. Using the linear predictor coefficients and the cross-correlation coefficients of frequencies, the emotion classification test for four emotions, such as anger, sad, joy, and relaxation is performed with an artificial neural network. The results of the two parameters showed that the linear prediction coefficients have produced the better results for emotion classification than the cross-correlation coefficients of FFT spectra.

• The Journal of Korea Institute of Information, Electronics, and Communication Technology
• /
• v.3 no.4
• /
• pp.23-27
• /
• 2010

This paper performs emotion classification test to find out the best parameter of electroencyphalogram(EEG) signal. Linear predictor coefficients, band cross-correlation coefficients of fast Fourier transform(FFT) and autoregressive model spectra are used as the parameters of 10-channel EEG signal. A multi-layer neural network is used as the pattern classifier. Four emotions for relaxation, joy, sadness, irritation are induced by four university students of an acting circle. Electrode positions are Fp1, Fp2, F3, F4, T3, T4, P3, P4, O1, O2. As a result, the Linear predictor coefficients showed the best performance.

• 제어로봇시스템학회:학술대회논문집
• /
• 2002.10a
• /
• pp.75.5-75
• /
• 2002

$\textbullet$ FFT $\textbullet$ Autocorrelation $\textbullet$ Levinson_Durbin resolution $\textbullet$ LP coefficients $\textbullet$ LP cepstral Coefficients $\textbullet$ Incremental Learning

• Journal of the Institute of Electronics Engineers of Korea SD
• /
• v.40 no.12
• /
• pp.87-94
• /
• 2003

In this paper, we propose an IFFT/FFT design method to minimize quantization error in IEEE 802.11a WLAN. In the proposed algorithm, the twiddle coefficient of IFFT/FFT processor is manipulated by the statistics distribution of the input data at each stage. We applies this algorithm to radix-2/$^2$ SDF architecture. Both IFFT and FFT processor shares the circuit blocks cause to the symmetric architecture. The maximum quantization error with the 10 bits length of the input and output data is 0.0021 in IFFT and FFT that has a self-loop structure with the proposed method. As a result, the proposed architecture saves 3bits for the data to keep the same resolution compared with the conventional method.