Study On The Segmentals In Forensic Speaker Recognition Between The Shanxi Dialect And The Mandarin Chinese

Recently, Forensic Speaker Recognition has been applied in legal cases which involve human speech recordings, and has become a key method in crime investigation. However, factors like anti-detection capability and cross-language users make Forensic Speaker Recognition still challenging. For example, suspects might speak a different language in the disputed sample that is collected in a crime scene than in the know sample which is collected in the police station. In these cases, cross-language comparison has to be carried. Until now, the words, recognition experts can use, are firstly limited to identical utterances with the same Context of utterance, followed by the isolated phrases or the syllables with the same onset, final and tone, at the end or beginning of the sentences.Due to geographical and historical factors, the differences between Shanxi dialect and Mandarin Chinese are far larger than that between Guanhua and Mandarin Chinese. When it comes to Forensic Speaker Recognition between Shanxi dialect and Mandarin Chinese, it is difficult to obtain ideal and enough speech sample to compare with. It is then worth investigating other segmentals which could be used in Forensic Speaker Recognition.Literature comparison, Field working, acoustic phonetics and Statistical analysis are applied for current study. It aims to investigate the feasibility of Automatic Speaker Recognition between Shanxi dialect and Mandarin Chinese, the differences between Shanxi dialect and Mandarin Chinese, the feasibility of Speaker Recognition between creaky syllables in Shanxi Fencheng dialect and Mandarin Chinese, the feasibility of Speaker Recognition using identical syllables only differ in tones, identical syllables only differ in consonant and identical later loud vowel of different later diphtongs. Furthermore, non-phonetic information was also investigated to exanimate to what extent could they be used in Forensic Speaker Recognition. Certain findings are listed below,1. Results of Forensic Speaker Recognition that are obtained by Automatic Speaker Recognition within the same language is inaccurate, not mention speech comparison between Shanxi dialect and Mandarin Chinese. Forensic Speaker Recognition works between Shanxi dialect and Mandarin Chinese still need to be carried out by, or at least with the assistant of, speech experts.2. Shanxi dialect has different vowel sets from Mandarin Chinese, especially in areas like Jiao Cheng (5), Ping Lu (6), Guang Ling (12), Yang Cheng (6), Wan Rong (11), Jing Le (6) and Zhong Yang (6). It is then worth further investigating Shanxi dialect among these areas.3. When carry out Forensic Speaker Recognitionn between Shanxi dialect and Mandarin Chinese, possible segmentation order could be the following:syllable carries identical consonant, identical vowel and identical tone â†' identical syllables only differ in tones â†' identical syllables only differ in consonant which is caused by phonation type â†' identical syllables differs in consonant and tone which is caused by phonation type â†' identical syllables differ in consonant which is caused by adjacent articulator â†' identical syllables differ in consonant and tone which is caused by adjacent articulator â†' identical syllables differ in consonant which is caused by non-adjacent articulator â†' identical later loud vowel of different later diphtongs with same toneâ†'identical later loud vowel of different later loud diphtongs with different tones. However, certain concern needs to be taken into account when choose these segmentals.4. Non-phonetic personal characters which are observed from suspects could also be used in Forensic Speaker Recognitionn.These findings further contribute to the theory in Chinese Forensic Speaker Recognitionn, and it could possibly be used in real cases. Furthermore, findings from current study could also be used in Forensic Speaker Recognitionn between Mandarin Chinese and other Chinese dialects.