Segmentation and modeling in segment-based recognition

Chang, Jane W.; Glass, James

doi:10.21437/eurospeech.1997-23

Cited by 18 publications

(1 citation statement)

References 6 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Accuracy on NIST core test set (%) Robinson, 1994 [6] 73.9 Deng and Sameti, 1996 [7] 73. 4 Mari et al, 1996, [8] 68.8 Chang and Glass, 1997, [9] 73.4 Ming and Smith, 1998, [5] 74.4 Karnjanadecha and Zahorian 72.6…”

Section: Systemmentioning

confidence: 99%

An investigation of variable block length methods for calculation of spectral/temporal features for automatic speech recognition

Karnjanadecha¹,

Zahorian²

2000

6th International Conference on Spoken Language Processing (ICSLP 2000)

View full text Add to dashboard Cite

This paper presents an investigation of non-uniform time sampling methods for spectral/temporal feature extraction for use in automatic speech recognition. In most current methods for signal modeling of speech information, "dynamic" features are determined from frame-based parameters using a fixed time sampling, i.e., fixed block length and fixed block spacing. This work explores new methods in which block length and/or block spacing are variable. Three methods are suggested and each was tested with the TIMIT database using a standard HMM recognizer. Phone recognition experiments were conducted using the standard 39 phone set. The methods were also evaluated with various HMM model complexities. Experimental results indicated that none of the proposed nonuniform feature time sampling methods perform significantly better than fixed time sampling methods. However, the best results obtained with the front end are comparable to those obtained with current state-of-the-art systems. Also the performance of our monophone system surpasses that of most reported context-dependent monophone systems.

show abstract

Section: Systemmentioning

confidence: 99%

An investigation of variable block length methods for calculation of spectral/temporal features for automatic speech recognition

Karnjanadecha¹,

Zahorian²

2000

6th International Conference on Spoken Language Processing (ICSLP 2000)

View full text Add to dashboard Cite

show abstract

Self-Similar Layered Hidden Markov Models

Adibi

Shen

2001

Principles of Data Mining and Knowledge Discovery

View full text Add to dashboard Cite

Hidden Markov Models (HMM) have proven to be useful in a variety of real world applications where considerations for uncertainty are crucial. Such an advantage can be more leveraged if HMM can be scaled up to deal with complex problems. In this paper, we introduce, analyze and demonstrate Self-Similar Layered HMM (SSLHMM), for a certain group of complex problems which show self-similar property, and exploit this property to reduce the complexity of model construction. We show how the embedded knowledge of selfsimilar structure can be used to reduce the complexity of learning and increase the accuracy of the learned model. Moreover, we introduce three different types of self-similarity in SSLHMM, and investigate their performance in the context of synthetic data and real-world network databases. We show that SSLHMM has several advantages comparing to conventional HMM techniques and it is more efficient and accurate than one-step, flat method for model construction.

show abstract