My Projects

Lombard Effect in Alaryngeal Speakers

For late-stage laryngeal cancer patients, when other invention options fail, total laryngectomy is often used as the last resort to remove the pathological larynx (voice box). Due to the loss of their phonatory apparatus during the procedure, they need to learn to use an alternative speaking method (known as alaryngeal phonation) in order to regain verbal communication. To update, four alaryngeal speaking methods are available: esophageal speech, tracheoesophageal speech, electrolaryngeal speech, and pneumatic artificial laryngeal speech.

Research revealed that, regardless of type, alaryngeal speech are associated with reduced intelligibility and, particularly for esophageal and electrolaryngeal speech, diminished loudness. Patients’ ability to speak more loudly becomes an important aspect of speech rehabilitation.

Imagine talking with someone in a noisy cocktail party. It is very natural that you will automatically raise your voice in order to be heard by your friend(s). This process is known as the Lombard effect. This is unconscious and done unnoticed, and how much you raise your voice seems to depend on the level of the background noise. That is, the louder the background is, the more loudly you will speak.

The present study attempts to make use of the Lombard effect to help alaryngeal speakers speak more loudly. We measured their vocal output when they are exposed to loud background noise at 60, 65, 70, 75, 80, 85, 90, 95, and 100 dB SPL (white noise). Preliminary findings indicate that they are able to raise their vocal output by varying amounts.

It is expected that, in the future, this strategy can be used to enhance their vocal output and used as a biofeedback to improve their intelligibility.

Respiratory Muscle Strength Training (RMST) in Professional Singers

Breath support is considered a crucial component in singing. Yet, the physiology of singing is quite different from ordinary (tidal) breathing at rest. The difference mainly originates from the different respiratory levels required by tidal breathing and singing. Whereas tidal breathing requires one to inhale up to about 40% of his/her vital capacity, it is not uncommon for a singer to inhale beyond 70% of his/her vital capacity. Breath support is apparently important for professional singers; their singing performance could be improved greatly by strengthening their breath support. This could be achieved by strengthening respiratory muscles through respiratory muscle strength training (RMST).

Despite the number of previous studies examining effect of such training on respiratory support, effects of RMST on singers are still inconclusive. To address this knowledge gap, the study attempts to evaluate the true effect of RMST on singing. Specifically, it aims to investigate: (1) the immediate effect of RMST on singers in terms of pulmonary, aerodynamic and acoustic measures right after the training; and (2) the possible maintenance effect of RMST on general singers in terms of pulmonary, aerodynamic and acoustic measures one month after the training.

Cantonese English accent reduction and native-like accent enhancement characterization and training for primary school students and English teachers.
港式英語口音如何改善？小學學生與英語老師英語訓練的聲學與感知研究”

Accent refers to errors or articulatory deviations observed when producing sounds of the foreign language (L2). In general, the greater is the deviation in articulation of L2 sounds, the more is the observed accent. Hong Kong is an international city, and for a native Cantonese speaker, being able to speak English as a second language has many obvious advantages. Yet, it has been observed that many local Hong Kongers are speaking English with a varying degrees of accent. Our previous research studies have indicated that such errors in producing L2 English sounds largely originate from Cantonese – the speakers’ mother tongue.

This project aims at characterizing the phonological gap between the best achievable accent of HKCE by primary school students as represented by that of the best students and local English teachers, and that of native British RP English accent, by means of perceptual evaluation and acoustical analysis, both under a native British English and a Cantonese framework. By measuring some speech physics parameters, we will gather quantitative evidence to experiment with our pilot training for students and teachers on how to reduce their Cantonese English accent and enhance it to one which is more native-like. We will construct a web platform to disseminate the analysis results and quantitative measurement of monitoring of the effects and progress of training.

Vowel Production by Patients with Parkinson’s Disease

As many as 7 million individuals in the world are diagnosed with different degrees of Parkinson’s disease (PD). Among them, about 80% exhibit speech problems, affecting their respiration, articulation, prosody, and resonance during speech production, which results in hypokinetic dysarthria. While motor impairment associated with PD can be greatly managed by PD medication, hypokinetic dysarthria is largely unresponsive. PD speech tends to deteriorate with time, resulting in progressive worsening of verbal communication with reduced intelligibility that negatively impacts patients’ quality of life.

The reduced speech intelligibility in PD individuals has been attributed mainly to their incorrect vowel articulation, even at the beginning of PD onset. Recent kinematic studies revealed reduced amplitude and velocity of jaw, lip and tongue movements in PD during speech production. This corresponds to the vocal tract rigidity and hypokinesia of articulatory movements typically associated with PD. An accurate description of articulatory gestures can reveal important information regarding the nature and extent of articulatory deviation, and thus offer the best therapeutic intervention for PD individuals.

The study attempts to examine the vowel articulation by PD patients of Cantonese of different severities. In addition to traditional formant measures, an array of derived formant parameters will be used to reveal the actual articulatory gesture and tongue position during speech production (see Table 1). Results from the study should help develop a customized therapeutic regimen for PDs by providing an individualized treatment targeting at client-specific articulatory errors with the use of biofeedback.