Performance of Speech Recognition Algorithms in Musical Speech used for Speech-Language Pathology Rehabilitation
Bibliographic record
Abstract
Musical speech in speech-language pathology rehabilitation is the production of speech following simple musical (rhythmic or melodic) patterns. This type of speech is used to facilitate speech processing in patients. In this study, we examined the performance of current automatic speech recognition (ASR) algorithms in recognizing normal and musical speech. From a first list of 28 identified algorithms, 24 were excluded for reasons such as low accuracy rate, high computational cost, high price, difficulty of use, long runtime, implementation problems. The four algorithms included were those from Amazon Web Services (AWS Transcribe), Google Speech Recognition, IBM Watson and Rev AI. We ran the selected algorithms on 60 sentences recorded under four speech conditions (Melodic; Rhythmic; Regular Slow; and Regular Normal). All algorithms did perfectly in recognizing the normal speech. The two algorithms with the best performance in musical speech (rhythmic and melodic speech) were AWS Transcribe and IBM Watson, both providing recognition accuracy above 98%. When adding moderate level of white noise and reverberation to the stimuli, AWS Transcribe remained with an acceptable (> 70%) or satisfactory (> 95%) ASR performance. These results may guide the development of software that use ASR to enable patients to undergo self-directed sessions of music-based speech-language rehabilitation, such as the melodic intonation therapy for post-stroke aphasia. The possibility to recognize musical speech allows to compare a patient’s performance to corresponding target phrases and provide feedback in the absence of a clinician. Given the recommended high intensity of treatment and the limited availability of speech-language pathologists, such software would be highly valuable to our healthcare systems.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.001 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".