• Proceedings of the Korea Information Processing Society Conference
• /
• 2001.10a
• /
• pp.801-804
• /
• 2001

This article is concerned with automatic segmentation of two adjacent vowels for speech signals. All kinds of transition case of adjacent vowels can be characterized by spectrogram. Firstly the voiced-speech is extracted by the histogram analysis of vowel indicator which consists of wavelet low pass components. Secondly given phonetic transcription and transition pattern spectrogram, the voiced-speech portion which has consecutive vowels automatically segmented by the template matching. The cross-correlation function is adapted as a template matching method and the modified correlation coefficient is calculated for all frames. The largest value on the modified correlation coefficient series indicates the boundary of two consecutive vowel sounds. The experiment is performed for 154 vowel transition sets. The 154 spectrogram templates are gathered from 154 words(PRW Speech DB) and the 161 test words(PBW Speech DB) which are uttered by 5 speakers were tested. The experimental result shows the validity of the method.

• Phonetics and Speech Sciences
• /
• v.13 no.3
• /
• pp.71-78
• /
• 2021

Most neural-network-based speech synthesis models utilize neural vocoders to convert mel-scaled spectrograms into high-quality, human-like voices. However, neural vocoders combined with mel-scaled spectrogram prediction models demand considerable computer memory and time during the training phase and are subject to slow inference speeds in an environment where GPU is not used. This problem does not arise in linear spectrogram prediction models, as they do not use neural vocoders, but these models suffer from low voice quality. As a solution, this paper proposes a Tacotron 2 and Transformer-based linear spectrogram prediction model that produces high-quality speech and does not use neural vocoders. Experiments suggest that this model can serve as the foundation of a high-quality text-to-speech model with fast inference speed.

• The Transactions of the Korea Information Processing Society
• /
• v.13 no.6
• /
• pp.284-290
• /
• 2024

Speech emotion recognition (SER) is a technique that is used to analyze the speaker's voice patterns, including vibration, intensity, and tone, to determine their emotional state. There has been an increase in interest in artificial intelligence (AI) techniques, which are now widely used in medicine, education, industry, and the military. Nevertheless, existing researchers have attained impressive results by utilizing acted-out speech from skilled actors in a controlled environment for various scenarios. In particular, there is a mismatch between acted and spontaneous speech since acted speech includes more explicit emotional expressions than spontaneous speech. For this reason, spontaneous speech-emotion recognition remains a challenging task. This paper aims to conduct emotion recognition and improve performance using spontaneous speech data. To this end, we implement deep learning-based speech emotion recognition using the VGG (Visual Geometry Group) after converting 1-dimensional audio signals into a 2-dimensional spectrogram image. The experimental evaluations are performed on the Korean spontaneous emotional speech database from AI-Hub, consisting of 7 emotions, i.e., joy, love, anger, fear, sadness, surprise, and neutral. As a result, we achieved an average accuracy of 83.5% and 73.0% for adults and young people using a time-frequency 2-dimension spectrogram, respectively. In conclusion, our findings demonstrated that the suggested framework outperformed current state-of-the-art techniques for spontaneous speech and showed a promising performance despite the difficulty in quantifying spontaneous speech emotional expression.

• Journal of the Korean Society of Laryngology, Phoniatrics and Logopedics
• /
• v.19 no.2
• /
• pp.101-106
• /
• 2008

The speech signal has been put into a form suitable for storage and analysis by computer, several different operation can be performed. Filtering, sampling and quantization are the basic operation in digiting a speech signal. The waveform can be displayed, measured and even edited, and spectra can be computed using methods such as the Fast Fourier Transform (FFT), Linear predictive Coding (LPC), Cepstrum and filtering. The digitized signal also can be used to generate spectrograms. The spectrograph provide major advantages to the study of speech. So, author introduces the basic techniques for the acoustic recording, digital signal processing and the principles of spectrum and spectrogram.

• The Journal of the Acoustical Society of Korea
• /
• v.27 no.8
• /
• pp.418-426
• /
• 2008

In this paper, we propose an algorithm for the frequency channel segmentation using peaks and valleys in spectrogram. The frequency channel segments means that local groups of channels in frequency domain that could be arisen from the same sound source. The proposed algorithm is based on the smoothed spectrum of the input sound. Peaks and valleys in the smoothed spectrum are used to determine centers and boundaries of segments, respectively. To evaluate a suitableness of the proposed segmentation algorithm before that the grouping stage is applied, we compare the synthesized results using ideal mask with that of proposed algorithm. Simulations are performed with mixed speech signals with narrow band noises, wide band noises and other speech signals.

• Speech Sciences
• /
• v.13 no.3
• /
• pp.49-65
• /
• 2006

Recently, several theories on the digital speech signal processing expanded the communication boundary between human beings and machines drastically. The aim of this study is to collect dialectal speech in Korea on a large scale and to establish a digital speech data base in order to provide the data base for further research on the Korean dialectal and the creation of value-added network. 528 informants across the country participated in this study. Acoustic characteristics of vowels and consonants are analyzed by Power spectrum and Spectrogram of CSL. Test words were made on the picture cards and letter cards which contained each vowel and each consonant in the initial position of words. Plot formants were depicted on a vowel chart and transitions of diphthongs were compared according to dialectal speech. Spectral times, VOT, VD, and TD were measured on a Spectrogram for stop consonants, and fricative frequency, intensity, and lateral formants (LF1, LF2, LF3) for fricative consonants. Nasal formants (NF1, NF2, NF3) were analyzed for different nasalities of nasal consonants. The acoustic characteristics of dialectal speech showed that young generation speakers did not show distinction between close-mid /e/ and open-mid$/\epsilon/$. The diphthongs /we/ and /wj/ showed simple vowels or diphthongs depending to dialect speech. The sibilant sound /s/ showed the aspiration preceded to fricative noise. Lateral /l/ realized variant /r/ in Kyungsang dialectal speech. The duration of nasal consonants in Chungchong dialectal speech were the longest among the dialects.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.19 no.4
• /
• pp.685-692
• /
• 2024

In this paper, we propose a new CNN model to improve the recognition performance of command voice signals. This method obtains a spectrogram image after performing a short-time Fourier transform (STFT) of the input signal and improves command recognition performance through supervised learning using a CNN model. After Fourier transforming the input signal for each short-time section, a spectrogram image is obtained and multi-classification learning is performed using a CNN deep learning model. This effectively classifies commands by converting the time domain voice signal to the frequency domain to express the characteristics well and performing deep learning training using the spectrogram image for the conversion parameters. To verify the performance of the speech recognition system proposed in this study, a simulation program using Tensorflow and Keras libraries was created and a simulation experiment was performed. As a result of the experiment, it was confirmed that an accuracy of 92.5% could be obtained using the proposed deep learning algorithm.

• Speech Sciences
• /
• v.9 no.4
• /
• pp.237-252
• /
• 2002

MIT has been applied for nonfluent aphasic patients on the basis of lateralization of brain hemisphere. However, its applications for different languages have some inquiry for aphasic patients because of prosodic and rhythmic differences. The purpose of this study is to develop the Korean Melodic Intonation Therapy program using personal computer and its clinical effects for nonfluent aphasic patients. The algorithm was composed to voice analog signal, PCM, AMDF, Short-time autocorrelation function and center clipping. The main menu contains pitch, waveform, sound intensity and speech files on window. Aphasic patients' intonation patterns overlay on selected kMIT patterns. Three aphasic patients with or without kMIT training participated in this study. Four affirmative sentences and two interrogative sentences were uttered on CSL by stimulus of ST. VOT, VD, Hold and TD were measured on Spectrogram. In addition, articulation disorders and intonation patterns were evaluated objectively on spectrogram. The results indicated that nonfluent aphasic patients with kMIT training group showed some clinical effects of speech intelligibility based on VOT, TD values, articulation evaluation and prosodic pattern changes.

• Asia-pacific Journal of Multimedia Services Convergent with Art, Humanities, and Sociology
• /
• v.7 no.6
• /
• pp.239-246
• /
• 2017

This paper compares the speech signal parameters between Korean and Chinese for Korean pronunciation /ㄹ/, which is caused many errors by Chinese leaners. Allophones of /ㄹ/ in Korean is divided into lateral group and tap group. It has been investigated the reasons for these errors by studying the similarity and the differences between Korean /ㄹ/ pronunciation and its corresponding Chinese pronunciation. In this paper, for the purpose of comparison the speech signal parameters such as energy, waveform in time domain, spectrogram in frequency domain, pitch based on ACF, Formant frequencies are used. From the phonological perspective the speech signal parameters such as signal energy, a waveform in the time domain, a spectrogram in the frequency domain, the pitch (F0) based on autocorrelation function (ACF), Formant frequencies (f1, f2, f3, and f4) are measured and compared. The data, which are composed of the group of Korean words by through a philological investigation, are used and simulated in this paper. According to the simulation results of the energy and spectrogram, there are meaningful differences between Korean native speakers and Chinese leaners for Korean /ㄹ/ pronunciation. The simulation results also show some differences even other parameters. It could be expected that Chinese learners are able to reduce the errors considerably by exploiting the parameters used in this paper.

• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.59 no.6
• /
• pp.1120-1125
• /
• 2010

In the speech system, for example hearing aid as well as speech communication, speech quality is degraded by environmental noise. In this study, to enhance the speech quality which is degraded by environmental speech, we proposed an algorithm to reduce the noise and reinforce the speech. The minima controlled recursive averaging (MCRA) algorithm is used to estimate the noise spectrum and spectral weighting factor is used to reduce the noise. And partial masking effect which is one of the human hearing properties is introduced to reinforce the speech. Then we compared the waveform, spectrogram, Perceptual Evaluation of Speech Quality (PESQ) and segmental Signal to Noise Ratio (segSNR) between original speech, noisy speech, noise reduced speech and enhanced speech by proposed method. As a result, enhanced speech by proposed method is reinforced in high frequency which is degraded by noise, and PESQ, segSNR is enhanced. It means that the speech quality is enhanced.