...
首页> 外文期刊>Cochlear implants international >Cantonese lexical tone recognition from frequency-specific temporal envelope and periodicity components in the same versus different noise band carriers
【24h】

Cantonese lexical tone recognition from frequency-specific temporal envelope and periodicity components in the same versus different noise band carriers

机译:从相同和不同噪声频带载波中特定于频率的时间包络和周期性分量中的粤语词汇音调识别

获取原文
获取原文并翻译 | 示例
           

摘要

Temporal envelope and periodicity components (TEPCs) in the speech signal offer important cues for speech recognition. A recent study revealed that TEPCs extracted from the high-frequency regions contribute significantly to Cantonese lexical tone recognition at the monosyllabic word level (Yuen et al., 2006). The aim of the current study was to compare the contributions of TEPCs from different frequency regions for Cantonese lexical tone recognition in connected speech, using the same or different noise band carriers and a very limited number of frequency bands. Test stimuli were Cantonese disyllabic word minimal pairs differing from each other only in lexical tone. The test stimuli set exhausted all possible 15 lexical tone-pair contrasts (six lexical tones in total) for the first and the second syllables. The TEPC of the test items were extracted by full-wave rectification and low-pass filtering and were used to modulate a speech-spectrum-shaped noise carrier to create the test stimuli. The stimuli therefore contain only TEPCs but no fine structure components (FSCs) of the original signal. Multiple sets of stimuli were created with different combinations of TEPC-modulated frequency bands. Twenty normal-hearing subjects participated in the study. Even in a single-band condition, much robust lexical tone recognition was obtained with TEPC provided from the high-frequency region than those from the low frequency region. This high frequency region TEPC advantage was consistent using noise band carriers from different frequency regions. The TEPCs in the high-frequency region are more important for Cantonese lexical tone recognition, when FSCs are not available. The results indicate a potential to improve speech recognition in tonal languages by preserving or enhancing the critical TEPCs via new signal'processing algorithms in hearing prosthesis,
机译:语音信号中的时间包络和周期性分量(TEPC)为语音识别提供了重要线索。最近的一项研究表明,从高频区域提取的TEPC在单音节单词级别上对粤语词汇音识别有显着贡献(Yuen等,2006)。当前研究的目的是使用相同或不同的噪声频带载波和非常有限的频带数目,比较来自不同频率区域的TEPC对粤语词汇音识别的贡献。测试刺激是粤语双音节最小对,仅词法上彼此不同。测试刺激集耗尽了第一个和第二个音节的所有可能的15个词法对对(总共六个词法对)。通过全波整流和低通滤波提取测试项目的TEPC,并将其用于调制语音频谱形状的噪声载体以创建测试刺激。因此,刺激仅包含TEPC,而没有原始信号的精细结构成分(FSC)。用TEPC调制的频段的不同组合创建了多组刺激。 20名听力正常的受试者参加了研究。即使在单频带条件下,使用高频区域提供的TEPC也会比低频区域提供更强大的词汇音调识别。使用来自不同频率区域的噪声频带载波,这种高频区域TEPC的优势是一致的。当没有FSC时,高频区域中的TEPC对于粤语词汇音识别更为重要。结果表明,通过在人工假肢中使用新的信号处理算法来保留或增强关键的TEPC,可以改善音色语言的语音识别能力,

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号