The Use of Voice Recognition for Forensic Investigations - Coursework Example

? Forensic Phonetics Preliminary assessments that should be carried out on forensic analysis speech samples to establish their adequacy for speaker comparison and identification analysis The use of voice recognition for forensic investigations goes back to several centuries back with the first case to employ the use of voice recognition in a case being the trial of William Hulet in 1660. The use of voice recognition in forensic investigation is employed in cases where the witness in a particular case did not see the criminal but rather heard their voice while carrying out the crime. The witnesses’ correct recognition and recollection of the voice off the criminal will therefore determine the conclusion of the investigation or case. The success of voice recognition is pegged on the ability of the witness to correctly remember and identify without doubt the voice of the perpetrator of the crime. This is however difficult as the voice of the criminal can change over time, the memory of the voice tends to decrease as time goes by, the memory of the witness might also be influenced by familiar voices among other factors. In certain cases such as the Hulet case, suspects have been wrongly accused due to the inability of the witness to correctly identify and recognize the voice of the criminal. It is for this reason that speech samples are employed during forensic investigations to try and increase the accuracy of voice recognition. These speech samples are collected at different periods of time to be used later in voice recognition in identification. In identifying which samples will be most effective for accurate voice recognition several assessments should be made (Neustein, 2012). For these samples to be effective the time the human voice takes to change has to be taken into consideration while collecting and using the samples. It is believed that while the human voice is slow to change the exact period it takes to change varies. For some it may take a year for others a few years to come. While collecting speech samples therefore the time taken for a voice to change will determine whether the sample will be effective or not. Studies however have shown that latency in time between recordings decreases the accuracy of voice recognition. This means that contemporary recordings will tend to record a higher chance of accurate identification as compared to latent recordings. The listener experience and similarity of the voices in speech samples also determines how effective they will be in voice recognition. More experienced listeners will tend to be more accurate in voice recognition as compared to those who are not. Accuracy is also higher where the voices in the speech samples are dissimilar (Neustein, 2012). The identification accuracy of a speech sample can also be influenced by sample duration and acoustic quality. Surprisingly studies have shown that the duration of a voice sample does not significantly influence voice identification or recognition. Samples whether long or short can achieve the same level of accuracy. Where however there is phonemic variation the longer sample will provide a more accurate chance for voice recognition. Acoustic quality unlike duration is more complex as it involves factors such as background noise and bandwidth. Background noise may include a radio playing, sudden speech, television sounds, traffic sounds, and crowds among others. The recording device on which the sample has been done may also have a bigger or smaller bandwidth. Some devices such as the telephone, mobile phones have a smaller bandwidth which may decrease clarity of the voice as compared to voice recorders, cameras among others (Erikson, nd.). Aural or acoustic assessment of a speech sample therefore involves a combination of linguistic judgments by the examiner listening to the sample as well as the use of acoustic measurements used by phoneticians to describe speech. Linguistic judgments include speech errors, pathological speech problems, idiosyncrasies, dialect, and foreign accents among others. Acoustic measurements on the other hand include fundamental frequency mean, standard deviation and speaking rate. The acoustic assessment process can thus be summarized into several steps; the first one is the listening of the speech samples where the duration and sound quality are explored. The second step is the transcription of the material which makes inconsistencies in the samples conveniently searchable. The third step is linguistic analysis in which speech samples are assessed from a linguistic angle such as dialect, grammar, accent, sociodialect among others. The fourth step is acoustic analysis which involves analysis of formants, fundamental frequency intonation patterns among others and finally a summary of the findings (Neustein, 2012). The significance of a speaker’s accent, vocal pitch and voice quality in establishing speaker identity in forensic speaker comparison analysis Voice quality is a significant aspect of speaker comparison. The most common acoustic voice quality descriptors useful for speaker comparison are jitter and shimmer. Jitter refers to the period to period variation in fundamental frequency while shimmer is the corresponding variation in amplitude. The measures of jitter have been successfully employed in distinguishing pathological and normal voices among speakers. Other voice descriptors include glottal pulse shape and glottal pulse spectrum. Voice creaks are also an aspect of voice quality used in speaker comparison. The degree of creak usually varies between different speakers thus differentiating speakers. Some speakers can for example have a leaky voice meaning their glottal folds do not close completely when they speak, others have a whispery voice which indicates the incomplete voicing of normally voiced segments, other have a hoarse voice thus irregular closing of the folds while others might have a falsetto which implies an increase in fundamental frequency resulting in a sharply falling glottal spectrum. Such aspects of voice quality can therefore enable the forensic investigator to accurately compare and distinguish different speakers (Neustein, 2012). A speaker’s accent refers to the speech sound traits of the speaker’s native language carried over to the second language. The speaker’s accent is an important aspect of speaker comparison as it is through this accent that the speaker suggests his or her culture, ethnicity, country or region of origin. The accent can also provide the investigator with the socioeconomic status of the speaker thus increasing the chances of speaker identification (Tanner, 2007). Finally the voice pitch is significant especially in the identification of gender during speaker comparison. The pitch level between the female and the male is often significantly large enough for the investigat9r to accurately distinguish between the male and female speaker. The pitch levels can also enable the investigator to compare the speech samples of the suspect or accurately identify the speaker in speaker comparison. The analysis of the voice pitch can also be used in speaker comparison where a voice disguise has been used to alter the voice of the speaker. In the case where there are difficulties in identifying the pitch such as where a male has a higher pitch than usual or female lower than usual formants or fundamental frequencies can be employed. The voice pitch of a speaker is however more often than not unique to the speaker thus a key aspect of speaker comparison (Muller, 2007). References Erikson, A., n.d. Forensic phonetics. Gothenburg: Department of Lingusitics Gothenberg University. Muller, C., 2007. Speaker classification 1: Fndamentals features and methods. London: Springer. Neustein, A., 2012. Forensic speaker recognition: Law enforcement and counter terrorism. London: Springer. Tanner, D. C., 2007. Medical -legaland Forensic aspects of communication disorders, voice prints and pseaker profiling. New York: Lawyer and Judges publishing company. Read More