The safety and scientific validity of this study is the responsibility of the study sponsor and investigators. Listing a study does not mean it has been endorsed by the ANZCTR. Before participating in a study, talk to your health care provider and refer to this information for consumers
Trial registered on ANZCTR

Registration number
Ethics application status
Date submitted
Date registered
Date last updated
Date data sharing statement initially provided
Type of registration
Retrospectively registered

Titles & IDs
Public title
Bionic voice reconstruction in dysphonic individuals
Scientific title
Assessment and validation of bionic voice reconstruction in dysphonic individuals
Secondary ID [1] 288336 0
Nil known
Universal Trial Number (UTN)
Trial acronym
Linked study record

Health condition
Health condition(s) or problem(s) studied:
Dysphonia 297311 0
Laryngectomee 297312 0
Aphonia 297313 0
Condition category
Condition code
Physical Medicine / Rehabilitation 297511 297511 0 0
Other physical medicine / rehabilitation
Neurological 297815 297815 0 0
Other neurological disorders

Study type
Patient registry
Target follow-up duration
Target follow-up type
Description of intervention(s) / exposure
Evaluation of novel computer algorithm that may reconstruct words from recorded partial speech or whispers. Subjects will be asked to speak phonetically balanced sentences for 15 minutes and this will be recorded. Then the computer algorithm will be used to reconstruct speech. The accuracy and rate of reconstruction of the regenerated voice will be rated by both researchers and the subjects. Ultimately the algorithm would be able to be used on a smart phone or device. A single visit only is required. Participants will be invited to hear reconstructed speech samples and rate them subjectively once reconstruction is complete. Files can be emailed to participants.
Intervention code [1] 293889 0
Not applicable
Comparator / control treatment
The participants are only required to provide voice samples. There is no randomisation or control group. Computer based regeneration is completed within a lab without a participant needing to be present.
Control group

Primary outcome [1] 297323 0
Accuracy of reconstructed voice samples compared to actual voiced sentences. Accuracy assessed by percentage correct measure where number of words correct is calculated vs actual spoken sentence.
Timepoint [1] 297323 0
Single session only
Secondary outcome [1] 320730 0
Rate of reconstructed voice samples compared to normal speech. This will be assessed by asking raters to classify samples as 'real time', 'slower than normal', 'very slow', 'unable to understand'.
Timepoint [1] 320730 0
Single session only

Key inclusion criteria
1) Abnormal voice
2) Able to attempt repeated voice tasks and follow instructions
3) Able to understand and give consent for the study and able to complete self-assessment questionnaire

Minimum age
18 Years
Maximum age
No limit
Both males and females
Can healthy volunteers participate?
Key exclusion criteria
1) Significant pulmonary disease condition
2) Unable to attempt voice tasks or follow instructions
3) Unable to speak English
4) Inability to give informed consent

Study design
Natural history
Defined population
Statistical methods / analysis
Data captured will be de-identified and then analysed through the Engineering Department at Unitec. Reproduced speech will be graded for intelligibility, accuracy of spoken words, volume, quality and speed of reproduction by two blinded researchers. Quantitative statistical analysis will be undertaken on the voice samples and ratings. Statistical analysis will be performed with p values below 0.05 considered significant.
Power Calculation
Based on a 50% difference in expected speech intelligibility between unaltered voices and voice-reconstructed voices, 65 subjects will be able to detect a real difference where power of 80% and significance levels of 0.05 are used. A mix of severity of voice dysfunction would be expected, from mild impairment to aphonia (total loss of voice). This sample size would not allow specific subgroup analysis but will indicate if further subgroup analysis might be useful in future research.
A total of 70 subjects was agreed so that we could aim to broadly recruit 20 subjects each with mild, moderate and severe voice disorders and at least 10 laryngectomee subjects, hence a cohort of 70 subjects.

Recruitment status
Date of first participant enrolment
Date of last participant enrolment
Date of last data collection
Sample size
Accrual to date
Recruitment outside Australia
Country [1] 7584 0
New Zealand
State/province [1] 7584 0

Funding & Sponsors
Funding source category [1] 292873 0
Name [1] 292873 0
Unitec Institute of Technology
Address [1] 292873 0
Unitec Institute of Technology
Private Bag 92025
Victoria St West
Auckland 1142
Country [1] 292873 0
New Zealand
Primary sponsor type
Unitec Institute of Technology
Unitec Institute of Technology
Private Bag 92025
Victoria St West
Auckland 1142
New Zealand
Secondary sponsor category [1] 291617 0
Name [1] 291617 0
Address [1] 291617 0
Country [1] 291617 0
Other collaborator category [1] 278816 0
Name [1] 278816 0
University of Auckland
Address [1] 278816 0
Private Bag 92019
Auckland 1142
Country [1] 278816 0
New Zealand
Other collaborator category [2] 278817 0
Name [2] 278817 0
North Shore Hospital
Address [2] 278817 0
Shakespeare Rd
Auckland 0740
Country [2] 278817 0
New Zealand

Ethics approval
Ethics application status
Ethics committee name [1] 294376 0
Northern B Committee
Ethics committee address [1] 294376 0
PO Box 5013
Wellington 6011
Ethics committee country [1] 294376 0
New Zealand
Date submitted for ethics approval [1] 294376 0
Approval date [1] 294376 0
Ethics approval number [1] 294376 0

Brief summary
This translational research proposal aims to develop and evaluate novel algorithms for creating a new rehabilitative method to restore natural sounding speech for those with significant dysphonia and laryngectomy patients, through a real-time speech processing framework. The new method would be non-invasive, non-surgical and capable of turning whisper-like voice into normal speech.
Participants in the study will provide voice samples which will be recorded in a sound-proofed room. These samples will be analysed using computing algorithms to enhance and 'fill in' missing parts of speech to reconstruct full words and sentences. The study will evaluate the accuracy of the reconstruction eg. did the computer regenerate the correct words? and the speed at which the reconstruction can be achieved. It will also evaluate the intelligibility of the reconstructed speech by blind ratings from both researchers and subjects.
Trial website
Trial related presentations / publications
Accepted for presentation at Combined Otolaryngological Societies Meeting, Chicago, USA (18 May 2016)
Public notes

Principal investigator
Name 54286 0
Dr Jacqueline Allen
Address 54286 0
Auckland Voice and Swallow
PO Box 99743
Auckland 1149
Waitemata District Healthboard work location
Country 54286 0
New Zealand
Phone 54286 0
Fax 54286 0
Email 54286 0
Contact person for public queries
Name 54287 0
Dr Jacqueline Allen
Address 54287 0
Auckland Voice and Swallow
PO Box 99743
Auckland 1149
Country 54287 0
New Zealand
Phone 54287 0
Fax 54287 0
Email 54287 0
Contact person for scientific queries
Name 54288 0
Dr Jacqueline Allen
Address 54288 0
Auckland Voice and Swallow
PO Box 99743
Auckland 1149
Country 54288 0
New Zealand
Phone 54288 0
Fax 54288 0
Email 54288 0

Data sharing statement
Will individual participant data (IPD) for this trial be available (including data dictionaries)?
No/undecided IPD sharing reason/comment
Trademark and patent applications are pending on the software analysis of the voice samples.
What supporting documents are/will be available?
Study protocol
Informed consent form
Summary results
No Results