Clinical pilot study assessment of a portable real-time voice analyser (Paper presented at PEVOC-IV)

Transcription

1 Batty, S.V., Howard, D.M., Garner, P.E., Turner, P., and White, A.D. (2002). Clinical pilot study assessment of a portable real-time voice analyser, Logopedics Phoniatrics Vocology, 27, Clinical pilot study assessment of a portable real-time voice analyser (Paper presented at PEVOC-IV) Sally V Batty (1), David M Howard (2), Paul E Garner (2), Pete Turner (2) and Andy D White (2) (1) Speech and Language Therapy, Community Health Sheffield, Sheffield, UK (2) Music Technology Research Group, Department of Electronics, The University of York, Heslington, York, YO10 5DD, UK. Abstract The use of computer-based real-time visual displays is now well established in voice clinics. A number of situations exist however, where a computer-based system is inconvenient and a battery-powered hand-held version would be preferred. This would, for example, enable voice monitoring to be carried out at any time and place, including during home visits. This paper describes a purpose-built device that provides real-time displays of fundamental frequency and larynx closed quotient derived from the electrolaryngograph output. The fundamental frequency data can in addition, be acquired from a microphone. The system has been in use recently in a speech and language therapy clinic in Sheffield, UK and the advantages and disadvantages noted during that trial are discussed. Keywords: real-time visual display, larynx closed quotient, fundamental frequency, voice 1

2 Introduction The electrolaryngograph (3) has been in use in voice clinics for many years and a wealth of data has been recorded in relation to a number of voice conditions. In addition, the real-time visual display of the electrolaryngograph output waveform (Lx) and voice fundamental frequency (f0) derived from it has proved most beneficial in a clinical setting. Statistical analyses of these and other parameters derived from Lx (1) can provide useful summaries of the current state of a subject s voice, enabling for example, the tracking of progress during therapeutic intervention. These displays and voice analysis techniques have also found application in regard to furthering our understanding of the professional voice. Other parameters can be readily derived from the Lx waveform and larynx closed quotient (CQ) is one in particular that has provided useful data in regard to quantifying the vocal changes that occur with professional singing or acting training (7, 9, 10, 11). At present, real-time displays of these parameters is only possible through the use of a desk-top computer system. This is not particularly convenient in many practical situations either in the clinic or professional voice studio. This paper describes a battery-powered hand-held real-time voice analysis system, which provides displays of f0 and CQ. The findings during trial of the system in the Royal Hallamshire Hospital in Sheffield, UK are discussed. System description The voice measurement system (2) is designed to provide real-time displays and objective assessment of voice related parameters in remote settings that are useful and relevant to clinical assessment and therapy. It is very important that the system is easy and convenient to use. This paper explores its clinical effectiveness following a pilot trial period in a voice clinic. The voice measurement system is designed to provide real-time displays of f0 and CQ derived from the Lx waveform as well as the Lx waveform itself. In addition, the f0 output can be derived from a microphone input by means of a peak-picking algorithm first designed for application in a real-time cochlear implant (5). The system itself can be hand-held since it is housed in a shaped case whose overall dimensions are 26cm (L) by 14.5cm (W) by 5.5cm (D). It should be noted that this housing is an off-the-shelf product for convenience. The unit is batterypowered and will operate for over 100 hours on 4 AA cells. The display has a 128 by 128 pixel resolution. The user interface has been designed to have the minimum number of user controls, and the main functions are accessed via on-screen menus by means of just four buttons: scroll up, scroll down, select/start, and stop. A microphone socket and an electrolaryngograph input socket are provided, there is a gain control associated with the microphone input and there is an on/off switch for the system. Figure 1: The main phases of the Lx waveform. (Tx: fundamental period; CP: closed phase; OP: open phase) The electrolaryngograph (1, 3) measures the electrical impedance between two electrodes placed at the level of the larynx externally on the throat of the subject, and its output waveform typically takes the shape shown in figure 1. There is greater vocal fold contact area indicated as the 2

3 current flowing between the electrodes increases (or the electrical impedance is lowered), and this enables the Lx waveform to be interpreted as shown in the figure. The shape of this waveform remains essentially constant during all voiced sounds. The time taken for each complete cycle can be measured reliably and accurately and this is the fundamental period (1, 8), indicted as Tx in the figure. Since f0 is defined as the number of periods completed in one second, it can be derived directly from Tx as follows: f0 = (1/Tx). Two other derived measurements are indicated in the figure: the closed and the open phase (1, 8), which indicate when the vocal folds are in contact and apart respectively. Larynx closed quotient (CQ) is defined as the percentage of each cycle for which the vocal folds remain in contact, and it is derived as the percentage that CP is of Tx in each cycle as follows: (CQ = (CP/Tx)*100%). A detailed description of the measurement of CP and OP in practice is given in (7). It is important to note that the Lx waveform provides no indication as to whether the vocal folds are completely closed during the interval referred to in figure 1 as the closed phase, since the Lx waveform only indicates variation in vocal fold contact area. The working definition of CQ used is therefore larynx closed quotient to indicate the electrolaryngographic origin of the measurement itself, which is entirely based on the Lx waveform as indicated in (7). The hand-held system also allows f0 to be measured from a microphone input. There is no one method for the measurement of f0 from the speech signal that gives reliable results for all speakers, all voice qualities and in any local acoustic. The choice of a suitable system has to be made by paying due consideration to the nature of the input speech signal that will typically be encountered as well as the errors that can be tolerated (4). A time domain peak-picking algorithm that was first implemented for use in single-channel cochlear implants is employed in the hand-held device (5). Its main advantage in this application is that it employs no output averaging and it will give an output even when the input speech has a rough or creaky voice quality. The only user setting is an input gain control that is used to ensure that the speech signal is of a suitable level for analysis. A detailed comparison between the peak-picker and other standard f0 analysis algorithms can be found in (8). Using the system There are two key menus in the system, voice analyser and SINGAD one each for displays and settings relating to the electrolaryngograph and microphone inputs. Their operation is illustrated in figure 2 with respect to the settings and displays that are currently available. The menu lists are navigated by means of the up and down arrow keys and then pressing the start/select button. The stop button is used to exit from display mode and to pause plotting when appropriate. The SINGAD menu provides control of the f0 plot, which was originally implemented as a singing pitching assessment display based on the SINGAD (SINGing Assessment and Development) system (6). In this mode, the frequency ( Set scale ) and time ( Timebase ) axes can be set by the user to preset values that have proved useful in previous versions (e.g. 6). 3

4 Figure 2: The settings, displays and two system menus available on the real-time voice analyser. When the electrolaryngograph is being used the autoset function provides a convenient way of setting up the system for an optimally sized Lx waveform plot. The user is asked to sustain a steady ah vowel for a few seconds. During this time the input gain is adjusted automatically to give a sensible amplitude to the displayed Lx waveform. In addition, the sampling rate is altered such that two cycles appear on the display for the sustained vowel. This enables a change in f0 of an octave above or below that used for the sustained vowel (when one cycle or four cycles will be displayed respectively), which provides a useful display in all normal situations. The CQ display plots values between 20% and 80% (CQ values are not typically observed outside this range). The f0 display is logarithmic and has the three settings available, octave ranges between 110Hz and 220Hz, 220Hz and 440Hz, and 440Hz and 880Hz. The system menu entries allow access to overall functions such as screen brightness and target shapes, and the Cx entry will provide in the future, a scatter plot of F0 as described in (1). Pilot study device assessment The system, which is illustrated in figure 3, has been used clinically on a regular basis as a pilot trial in the Speech and language therapy department of the Royal Hallamshire Hospital in Sheffield, UK. The purpose was to compare its use with existing computer-based electrolaryngograph system with specific attention to the ease of use, the accuracy of the displays and the overall potential usefulness and acceptability of the system in practice. 4

5 Figure 3: The device is shown connected to a Field Laryngograph (from Laryngograph Ltd.) and an inset showing the Lx waveform display with its associated CQ value. As a result of this trial period, the therapist involved noted a number of points. The unit was found to be easy to handle in terms of the user interface and the visibility of the display. The smallness of the unit was commented upon. In practice it was noted that the system was neither too complex nor over-powering for clients. The menus were seen as providing a simple operating procedure. The readings obtained from the system were found to be fully comparable with existing electrolaryngograph-based equipment. Of particular importance was the short setting up time in terms of the length of time required to unpack and get working compared to the existing system. The autoset feature was found to be very useful for registering new clients to enable a suitably sized Lx waveform display to be obtained. Overall comments made on reflection after the trials included the fact that the device takes up little space and therefore can be readily carried around. This was found to be of particular benefit since the system could be taken to the client and therefore the territory could be kept neutral as required. In particular, there is now the opportunity to make measurements in the client s home and changes in room availability would no longer be an issue. One client was extremely keen to take a unit home in order to practice and was asking for homework exercises between therapy sessions. A number of additional features were mentioned that would enhance the system s overall usefulness in practice. The screen size could be larger and a hard copy facility would be very useful. The data could be stored for future analysis and a display of statistical information would be beneficial. It was also suggested that an autoset feature for the f0 and CQ ranges that could be set for each client would be a useful addition that would also enhance the useful screen area. All autoset settings could then be stored on an individual basis for each client. Conclusions A hand-held voice analysis has been described. It is designed to operate in conjunction with the electrolaryngograph to provide displays of the output waveform, fundamental frequency and larynx closed quotient. It is controlled by means of two menus that are accessed using four buttons. A pilot study trial has indicated that the system has a number of advantages over existing computer-based equipment, mainly due to its small size, portability, ease of use, and acceptability with clients. The study clearly demonstrated the potential usefulness of such a system in practice as well as identifying a number of potentially very useful additions that could be made. All of these are technically feasible since the system is entirely software-based. A real implementation issue now exists though, as to whether it is appropriate to implement such a system using special purpose hardware given the ubiquity of hand-held personal organisers. These devices are mass produced and therefore highly cost effective, and they have larger display screens often with the option of colour. It is suggested that this could be the appropriate way ahead in the future. There are a number of technical issues to overcome however, most notably with respect to finding such a 5

6 device that incorporates a suitable audio input (typically there is either no audio capability or audio input is only possible via a microphone mounted in the case). Acknowledgements The authors thank the staff and clients at the Royal Hallamshire Hospital, Sheffield UK who took part in the pilot study as well as Dave Hunter and Mark Hough for their contributions to the development of the system. References 1. Abberton, E.R.M., Howard, D.M., and Fourcin, A.J. (1989). Laryngographic assessment of normal voice: A tutorial. Clinical Linguistics and Phonetics, 3, Batty, S.V., Garner, P.E., Howard D.M., Turner, P., and White, A.D. (2000). The development of a portable real-time display of voice source characteristics, Proceedings of the 26 th Euromicro Conference, Maastricht, 2, Fourcin, A.J, and Abberton, E.R.M. (1971) First applications of a new laryngograph, Medical and Biological Review, 21, Hess, W. (1983). Pitch determination of speech signals, Berlin: Springer-Verlag. 5. Howard, D.M., and Fourcin, A.J. (1983). Instantaneous voice period measurement for cochlear stimulation, Electronics Letters, 19, (19), Howard, D.M., and Welch, G.F. (1993). Visual displays for the assessment of vocal pitch matching development, Applied Acoustics, 39, (3), Howard, D.M. (1995). Variation of electrolaryngographically derived closed quotient for trained and untrained adult female singers, Journal of Voice, 9, Howard, D.M. (1998). Practical voice measurement, In: The voice clinic handbook, Harris, T., Harris, S., Rubin, J.S., and Howard, D.M. (Eds.), London: Whurr Publishing Company, Rossiter, D.P., Howard, D.M., and Comins, R. (1995). Objective measurement of voice source and acoustic output change with a short period of vocal tuition, Voice, 4, (1), Rossiter, D.P., Howard, D.M., and De Costa, M. (1996) Voice development under training with and without the influence of real-time visually presented biofeedback, Journal of the Acoustical Society of America, 99, (5), Rossiter, D.P., and Howard, D.M. (1998). Observed change in mean speaking voice fundamental frequency of two subjects undergoing voice training, Logopedics Phoniatrics Vocology, 22, (4),

7 7