An acoustic segment modeling approach to query-by-example spoken term detection

Wang, Haipeng; Leung, Cheung-Chi; Lee, Tan; Ma, Bin; Li, Haizhou

doi:10.1109/icassp.2012.6289081

Cited by 50 publications

(29 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Each ASM unit had 3 states with 16 gaussian components at each state. Detailed training procedure of the ASM tokenizer can be found in [9]. Both the GMM tokenizer and the ASM tokenizer took in 39-dimensional MFCC features, which were processed with utterance-based mean and variance normalization (MVN) and vocal tract length normalization (VTLN).…”

Section: Tokenizer Implementationmentioning

confidence: 99%

“…In the previous studies related to the posteriorgram-based template matching framework, many efforts were contributed to introducing novel modeling for the tokenizers, such as deep Boltzmann machine [7], discriminant GMM [8], acoustic segment model (ASM) [9], etc. However few works were dedicated to the combined use of different tokenizers.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Using parallel tokenizers with DTW matrix combination for low-resource spoken term detection

Wang

Lee

Leung

et al. 2013

2013 IEEE International Conference on Acoustics, Speech and Signal Processing

Self Cite

View full text Add to dashboard Cite

Recently the posteriorgram-based template matching framework has been successfully applied to query-by-example spoken term detection tasks for low-resource languages. This framework employs a tokenizer to derive posteriorgrams, and applies dynamic time warping (DTW) to the posteriorgrams to locate the possible occurrences of a query term. Based on this framework, we propose to improve the detection performance by using multiple tokenizers with DTW distance matrix combination. The proposed approach uses multiple tokenizers in parallel as the front-end to generate different posteriorgram representations, and combines the distance matrices of the different posteriorgrams into a single matrix. DTW detection is then applied to the combined distance matrix. Lastly score post-processing techniques including pseudo-relevance feedback and score normalization are used for further improvement. Experiments were conducted on the spoken web search datasets of MediaEval 2011 and MediaEval 2012. Experimental results show that combining multiple tokenizers significantly outperforms the best single tokenizer, and that the DTW matrix combination method consistently outperforms the score combination method when more than three tokenizers are involved. Score post-processing techniques show further gains on top of using multiple tokenizers.

show abstract

Section: Tokenizer Implementationmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Using parallel tokenizers with DTW matrix combination for low-resource spoken term detection

Wang

Lee

Leung

et al. 2013

2013 IEEE International Conference on Acoustics, Speech and Signal Processing

Self Cite

View full text Add to dashboard Cite

show abstract

“…However, the proposed framework differs from previous studies in that features derived from a pre-index of spoken documents are used to train a classifier. Most previous studies used acoustic features [19], [20], ASR-related information [21], latticebased information [22], [23], and similar approaches. In contrast, we demonstrate the effectiveness of the proposed decision process, wherein STD outputs are verified using an SVM-based classifier trained with pre-indexed best match keyword-based features.…”

Section: Introductionmentioning

confidence: 99%

Spoken Term Detection Using SVM-Based Classifier Trained with Pre-Indexed Keywords

Domoto

Utsuro

Sawada

et al. 2016

IEICE Trans. Inf. & Syst.

View full text Add to dashboard Cite

SUMMARYThis study presents a two-stage spoken term detection (STD) method that uses the same STD engine twice and a support vector machine (SVM)-based classifier to verify detected terms from the STD engine's output. In a front-end process, the STD engine is used to preindex target spoken documents from a keyword list built from an automatic speech recognition result. The STD result includes a set of keywords and their detection intervals (positions) in the spoken documents. For keywords having competitive intervals, we rank them based on the STD matching cost and select the one having the longest duration among competitive detections. The selected keywords are registered in the pre-index. They are then used to train an SVM-based classifier. In a query term search process, a query term is searched by the same STD engine, and the output candidates are verified by the SVM-based classifier. Our proposed twostage STD method with pre-indexing was evaluated using the NTCIR-10 SpokenDoc-2 STD task and it drastically outperformed the traditional STD method based on dynamic time warping and a confusion network-based index.

show abstract

“…The input of the GMM tokenizer was 39-dimensional MFCC feature vectors, which had been processed with vocal tract length normalization (VTLN). The second is an ASM tokenizer [14], containing 256 ASM units. Each unit had 3 states with 16 Gaussian mixtures for each state.…”

Section: Restricted Systemsmentioning

confidence: 99%

“…Five semi-supervised tokenizers were built from Czech, Hungarian, Russian, English and Mandarin phoneme recognizers, which were all in the split temporal context network structure [12,13]. The two unsupervised tokenizers were "GMM" and "ASM" (Acoustic Segment Modeling) [14], as described in Section 4.2. All these tokenizers were used to generate posteriorgrams, and Dynamic Time Warping (DTW) was applied for detection.…”

Section: Open Systemsmentioning

confidence: 99%