The Analysis Of Mandarin Tone Recongnition About Monosyllabic Word

BackgroundThere are more than 27.8 million deaf patients in the nation,on the basic of the second national sample survey of persons with disabilities in 2006.The number is still on the increase.Among the total,the deaf children who are from zero to seven years old add up to 800,000 each year,and there are 30 thousand new born deaf childrenwith the population growth and aging of life expectancy,the number of presbyacousis is increasing,reaching 9.49 million.The causes of the deafness are different.According to the region and the nature of the disease,deafness is divided into three categories:conductive deafness,sensorineural deafness and mixed deafness.Hearing impairment seriously affects their social interaction and personal quality of life.A good few of treatments do not get effectness,so till now it has no good methd.Patients who are irresponsive to the medicine treatments can only get help with the hearing aid or do the cochlear implant.Mandarin is a tone language,a total of four tones:the first,second,third and fourth tone respectively.A word is a syllable,each syllable has four tones.Chinese Tone,also known as word or syllable tone,are not only the composition of the words but also discriminating the meanings of words.It Mainly depends on the pitch (pitch frequency),different tone has various tone duration and tone intensity.Among the tone languages such as Chinese,the syllable tone is very important.If the vowel and consonant of the syllables are the same,the meaning is completely different for the different tone.The previous studies show that the information of acoustic perception on tone Recognition is widely distributed in frequency domain and time domain.Based on both the relative availability and the importance of time and frequency domain,they are mutual compensation each other.The most important information of the tone recognition is the changes in voice frequency,it manifests the change of the fundamental frequency and its harmonic components at the acoustic.The fundamental frequency F0 is the chief feature of tone recognition.The tone recognition is almost perfect,if either directly retain the fundamental frequency(F0) information or indirectly abstract the residual frequency from the harmonic structure through high-frequency filtering.Although F0 is the main feature,it is not the only information,the other information that can pass the time-domain model of mandarin tone also contributes to tone recognition, including the amplitude envelope,cyclical fluctuations,fine structure and vowel duration,and so on.When the F0 and its harmonic structure are removed partly and fully,vowel duration and amplitude envelope maintain the effective tone information,on the contrary,retaining the information of the fundamental frequency, their impact on Chinese Tone Perception is much smaller.There is a correlation between the amplitude envelope and F0 contour in Chinese syllable,the amplitude envelope plays a more significant role than the cyclical fluctuations on tone identification.The information of time-domain envelope mainly supports the speech recognition,but time-domain fine structure is major and necessary in the tone recognition.In a quiet condition,the tone recognition of the people with normal hearing is almost perfect only by the application of the time-domain fine structure, but only with the envelope time-domain information,the correct rate of the mandarin tone is relatively low,or so 70-80%;in the noise condition,the fine structure plays a more important role than the envelope information,envelope information is more sensitive to the noise in the tone perception.So that,if more fine structure information is provided for the stimulus in the cochlear implant,it is possible to improve tone recognition performance in patients Some researches have pointed out that in the quiet conditions,when a few number of channel in cochlear implant and the simulate frequency is higher,or the more channel in cochlear implant and the simulate frequency is lower,the tone recognition is similar to each other;in the noise condition,the cyclical information is more sensitive to noise,but the frequency-domain information can tolerate noise and play a major role in the tone recognition.So that it is presumed that there is a set-off compensation between the time domain and frequency domain information in Chinese Tone Recognition.Multi-channal artifical cochear implant has become an important means of dealing with the severe and extreme sensorineural deafness.At present,the multi-channel cochlear implant products used in the nation are majorly made in the foreign countres.It is suggested that its tone recognition is ineffective,because the design on its speech coding scheme is based on the characteristics of the Western language,and does not take into account the characteristics of Chinese speech. Whether the existing speech coding program is applicable to the Chinese or not,we need analyse the characteristics of Mandarin,in particular,constitution of the mandarin tone.So,in order to improve the tone recognition rate of cochlear implant users to provide further experimental basis,we have to analyse the time domain and frequency domain of the Mandarin tone,and design a new program extracting and coding the exact fundamental frequency and its harmonics.Objective To investigate the factors Affecting of Chinese speech tone recognition,and analyzes the Mandarin Monosyllabic word phoneme in both time and frequency domains.With the method of digital filtering,to analyzes the range of Speech frequency in the tone of the Mandarin Monosyllabic word.Methods1.MaterialsMaterials used in this study come from the book,named "the instruction manual of hearing and speech rehabilitation assessment for deaf children." This manual was printed and published in 1991 by audio-video Education Publishing House in Jilin Province.The part of auditory function assessment was read by female announcer who spoke the standard Putonghua(Mandarin)and recorded in a compact disk.We will take advantage of the part of the tone recognition about the monosyllabic word,a total of 10 syllables,four tones,adding up to 40 words.2.Experiment Process:Three main process:(1) the time-domain analysis The audio signals were extracted from the VCD video files and converted to.Wav files and stored in the disk,using the software of Cool Edit Pro 2.0 that was developed by the Syntrillium Software Company in the United States,audio sampling rate of 44100Hz,the sampling accuracy of 16 bits, stereo sound gate.The time-domain waveforms of the four tones in each monollabic word were displayed and their envelopes were extracted through the software of Cool Edit Pro 2.0,and the duration of the sounds were recorded by measurement tools.Study wheher each tone has its unique characteristics in the time-domain waveform,envelope,and the duration,to explore how the time-domain informations make a great impact on the tone recognition of Mandarin monosyllabic words.(2) the frequency-domain analysis The frequency domain analysis of the audio Mandarin Monosyllabic word phoneme were done with Fast Fourier transform(FFT) for amplitude spectra analysis.The collected audio datas of Mandarin Monosyllabic word were preprocessed and made a time-frequency analysis with the software of MatLab 7.0.The three-dimensional figures were plotted on the basic of the time-frequency analysis information using the SigmaPlot 9.0 software.Research the impact of fundamental frequency and formant frequency(frequency domain) information on the tone recognition of Mandarin monosyllabic words.(3) Filter analysis the above monosyllabic words were collected with Cool Edit Pro 2.0 software and transformed to audio files datas.The audio files datas were filtered with the Finite Impulse Respones(FIR) digital filter in high-pass at 500Hz, band-pass from 500Hz to 4000Hz and from 500Hz to 2000Hz,low-pass at 4000Hz and 2000Hz,respectively.The original and filtered speech signals were recognized by the six young college students with normal hearing,in order to distinguish the common meaning of the word.The frequency domain signals that were filtered with digital filter were transformed to the time domain signals,measuring the amplitude of time-domain waveform,to observe the effect of the different band-pass digital filters on the amplitude of the time-domain signal.The experimental datas of the first and second part were conducted Statistics analysed(all datas were described as mean±standard deviation,using SPSS13.0 to process datas and statistica analysed by relative methods) and Graphic manufacture(graphic analysis through the software of Cool Edit Pro 2.0,MatLab 7.0,SigmaPlot 9.0).Results1.The time-domain waveform and envelope of the monosyllabic words vary depending on its different tones.The time-domain analysis of waveform and temporal envelope are shown that regardless of the consonant and vowel, fundamental-frequency envelope is highly similar as long as the tone each other is the same in the different monosyllabic words.It is demonstrated that the time-domain information play an vital role in Chinese tone recognition.For the same monosyllabic word,its duration varied with the different tones and there is a significant difference for each other(P＜0.05).2.The frequency domain analysis indicates that monosyllabic words are mainly composed of F0,F1,F2 and F3,in which F0 is the fundamental frequency of the speech signals,F1 and F2 are the second harmonic and third harmonic.In particular, F3 is the high-frequency components of the speech signals and plays an important role in the speech clarity.The means of fundamental frequency in each group are statistically significant(P＜0.01),and there is extremely significant difference in each other(P＜0.01).The time-frequency-intensity-graphics indicates that the characteristics vary according to the tone of monosyllabic words.With the time change of the words,the various frequencies and its intensity of tonel remains basically unchanged,the curve is horizontal type;the different frequencies of tone2 gradually tend to high-frequency direction,the graph is upward-type,additionally, the intensity alters little;the divers frequencies of tone3 firstly deflect to low-frequency areas,then increasingly deviat to high-frequency direction of bias after maintaining a period of time,the figure is v-type,intensity changes are concave-shaped;the various frequencies of tone4 deviat from high-frequency to low-frequency areas,and its intensity reduces sharply,the graph is decreased type.3.Various band-pass filter results display that the amplittude of the waveform will reduce or the vioce of the monosyllabic word may change if the low-frequency components below 500Hz and/or the high-frequency components above 4000Hz are flitered out.There are good reasons to believe that the Chinese speech frequency coverage has exceeded the range from 500Hz to 4000Hz. Conclusion1.Time-domain information plays an important role in disicrimination of Mandarin tone about monosyllabic words.Through raising the sampling rate and stimulating rate of time domain in the cochlear implant system,it will provide more time datas and obtain more detailed time-domain waveform ang envelope information.2.Changes in the frequency information reflect the diversity of Mandarin Tone, in which fundamental frequency is major,but the formant frequencies to some extent,also provid the tone information.Frequency-domain information has an important role in Chinese Tone Recognition.Time-frequency analysis reflects the characteristics that the intensity and frequency change with the passage of time in the tone of Chinese monosyllabic words.3.The amplittude of the waveform degrades in different degrees and some words change in the voice following various digital filtering of the Chinese Monosyllabic word.It is sufficiently demonstrated that speech frequency is beyond the current range from 500Hz to 4000Hz.It is need further studed dose it is necessary to revise the frequency range of Chinese language.