The University College London Archive of Stuttered Speech (UCLASS)

An Automatic Prolongation Detection Approach in Continuous Speech With Robustness Against Speaking Rate Variations

In recent years, many methods have been introduced for supporting the diagnosis of stuttering for automatic detection of prolongation in the speech of people who stutter. However, less attention has been paid to treatment processes in which clients learn to speak more slowly. The aim of this study was to develop a method to help speech-language pathologists (SLPs) during diagnosis and treatment sessions. To this end, speech signals were initially parameterized to perceptual linear predictive (PLP) features. To detect the prolonged segments, the similarities between successive frames of speech signals were calculated based on correlation similarity measures. The segments were labeled as prolongation when the duration of highly similar successive frames exceeded a threshold specified by the speaking rate. The proposed method was evaluated by UCLASS and self-recorded Persian speech databases. The results were also compared with three high-performance studies in automatic prolongation detection. The best accuracies of prolongation detection were 99 and 97.1% for UCLASS and Persian databases, respectively. The proposed method also indicated promising robustness against artificial variation of speaking rate from 70 to 130% of normal speaking rate.

Past tense marking in the spontaneous speech of preschool children who do and do not stutter

PURPOSE

The aim of this study was to identify whether different patterns of errors exist in irregular past-tense verbs in children who stutter (CWS) and children who do not stutter (CWNS).

METHOD

Spontaneous language samples of thirty-one age- and gender-matched pairs of children (total N=62) between the ages of 24 months and 59 months were analyzed.

RESULTS

Results indicated that children who do and do not stutter over-regularize irregular past-tense verbs (i.e., saying runned for ran) with comparable frequency. However, two nonsignificant trends which suggest possible intra-group differences were noted. First, irregular past tense verbs represented a greater portion of total verbs for CWS than for CWNS. Second, CWS appeared to double-mark (i.e., say ranned for ran) more often than CWNS. Results are discussed in light of theories about the acquisition of the irregular past-tense and about differences in language skills between CWS and CWNS.

EDUCATIONAL OBJECTIVES

After reading this article, the reader will be able to: (a) summarize previous findings about connections between stuttering and language in CWS and CWNS; (b) describe similarities and differences between irregular past-tense verb use and errors in CWS and CWNS; (c) discuss possible connections between the declarative-procedural model and stuttering.

Comparison of alternative methods for obtaining severity scores of the speech of people who stutter

Riley's Stuttering Severity Instrument (SSI) is widely used. The manuals allow SSI assessments to be made in different ways (e.g. from digital recordings or whilst listening to speech live). Digital recordings allow segments to be selected and listened to, whereas the entire recording has to be judged when listened to live. Comparison was made between expert judges when they used these digital and live procedures to establish whether one method was more sensitive and reliable than the other. Five expert judges assessed eight speakers four times each in two judgment conditions (digital vs. live). The eight speakers were chosen so that they spanned a wide range of stuttering severity. SSI version 3 (SSI-3) estimates were obtained on all occasions. An ANOVA showed a three-way interaction between sessions, speakers and condition that indicated that digital and live judgments varied across speakers and across sessions. The predictions that were upheld were (1) SSI-3 scores made from digital segments are more sensitive than SSI-3 scores made on the entire live signal; (2) digital and live judgments vary with respect to speaker's stuttering severity and across test sessions.