Continuous hidden Markov modeling for speaker-independent word spotting

Rohlíček, Ján; Russell, William Howard; Roukos, Salim; Gish, H.

doi:10.1109/icassp.1989.266505

Cited by 197 publications

(99 citation statements)

References 2 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…If the area under the ROC is A upto a false alarm rate of f , then the percentage area under the ROC curve is given by P A ROC = The purpose of doing this normalization is to get rid of variations of area under the ROC because of differences in the support of the ROC curves. Another performance measure typically used to evaluate KWS performance is the Figure of Merit (FOM) [12] score, which is the mean of a modified ROC curve sampled at ten points. We have found P A ROC to be better than FOM in capturing gradual improvements of the model because FOM takes the value of the ROC curve at certain number of points and does not quantify the overall change (improvement or deterioration) of the curve.…”

Section: Experiments and Resultsmentioning

confidence: 99%

Low resource point process models for keyword spotting using unsupervised online learning

Sadhu

Ghosh

2017

2017 25th European Signal Processing Conference (EUSIPCO)

View full text Add to dashboard Cite

Abstract-Point Process Models (PPM) have been widely used for keyword spotting applications. Training these models typically requires a considerable number of keyword examples. In this work, we consider a scenario where very few keyword examples are available for training. The availability of a limited number of training examples results in a PPM with poorly learnt parameters. We propose an unsupervised online learning algorithm that starts from a poor PPM model and updates the PPM parameters using newly detected samples of the keyword in a corpus under consideration and uses the updated model for further keyword detection. We test our algorithm on eight keywords taken from the TIMIT database, the training set of which, on average, has 469 samples of each keyword. With an initial set of only five samples of a keyword (corresponds to ∼ 1% of the total number of samples) followed by the proposed online parameter updating throughout the entire TIMIT train set, the performance on the TIMIT test set using the final model is found to be comparable to that of a PPM trained with all the samples of the respective keyword available from the entire TIMIT train set.

show abstract

Section: Experiments and Resultsmentioning

confidence: 99%

Low resource point process models for keyword spotting using unsupervised online learning

Sadhu

Ghosh

2017

2017 25th European Signal Processing Conference (EUSIPCO)

View full text Add to dashboard Cite

show abstract

“…These factors also influence the reliability of detections and hence can be utilized in estimating the confidence of detections. Research has been conducted on confidence estimation utilizing various informative factors, in both automatic speech recognition and keyword spotting, e.g., Rohlicek et al (1989); Cox and Rose (1996); Bergen and Ward (1997); Kemp and Schaaf (1997); Ou et al (2001); Ayed et al (2002); Jiang (2005), and various methods have been employed to combine the heterogeneous informative factors, including decision trees (DT), general linear models (GLMs), generalized additive models (GAMs) and multi-layer perceptrons (MLPs) (Chase, 1997;Gillick et al, 1997;Zhang and Rudnicky, 2001). It has been found that features derived from multiple sources -with appropriate normalization -can be combined to serve as a good measure of confidence, which can in turn be used to evaluate the correctness of a recognition hypothesis or a keyword detection.…”

Section: Motivation and Organization Of This Papermentioning

confidence: 99%

“…For keyword spotting/spoken term detection, Rohlicek et al (1989) proposed to use duration-normalized acoustic likelihood as a confidence measure for each keyword in a filler model-based keyword spotting system, and Manos and Zue (1997) studied various features in a filler model-based keyword spotting system to compute the confidence score for each word, which included a segment phonemic match score, a score based on the probability of the particular segmentation, a lexical weight, a phone duration-based score, and a bigram transition score. Ou et al (2001) employed word posterior likelihoods derived from keyword, anti-keyword and non-keyword models and duration as features in a neural network classifier for utterance verification within a filler model-based keyword spotting system whereas Ayed et al (2002) used information based on the number of frames, the phone posterior probability, the frame-based phone posterior probability and the duration-based phone posterior probability along with the number of phones as lexical feature in an SVM classifier for utterance verification in an LVCSR-based keyword spotting system.…”

Section: Feature Collectionmentioning

confidence: 99%

Feature analysis for discriminative confidence estimation in spoken term detection

Tejedor

Toledano

Wang

et al. 2014

Computer Speech & Language

View full text Add to dashboard Cite

El acceso a la versión del editor puede requerir la suscripción del recurso Access to the published version may require subscription AbstractDiscriminative confidence based on multi-layer perceptrons (MLPs) and multiple features has shown significant advantage compared to the widely used lattice-based confidence in spoken term detection (STD). Although the MLP-based framework can handle any features derived from a multitude of sources, choosing all possible features may lead to over complex models and hence less generality. In this paper, we design an extensive set of features and analyze their contribution to STD individually and as a group. The main goal is to choose a small set of features that are sufficiently informative while keeping the model simple and generalizable. We employ two established models to conduct the analysis: one is linear regression which targets for the most relevant features and the other is logistic linear regression which targets for the most discriminative features. We find the most informative features are comprised of those derived from diverse sources (ASR decoding, duration and lexical properties) and the two models deliver highly consistent feature ranks. STD experiments on both English and Spanish data demonstrate significant performance gains with the proposed feature sets.

show abstract

“…While KWS is an active research area, most techniques are not suitable with our constraints. A common KWS approach is the Keyword/Filler Hidden Markov Model (HMM) [3,4,5,6,7]. It first builds a special decoding graph that contains both keywords and filler words, and then uses Viterbi decoding to determine the best path through the graph.…”

Section: Introductionmentioning

confidence: 99%

Query-by-example keyword spotting using long short-term memory networks

Chen

Parada

Sainath

2015

2015 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

133

View full text Add to dashboard Cite

We present a novel approach to query-by-example keyword spotting (KWS) using a long short-term memory (LSTM) recurrent neural network-based feature extractor. In our approach, we represent each keyword using a fixed-length feature vector obtained by running the keyword audio through a word-based LSTM acoustic model. We use the activations prior to the softmax layer of the LSTM as our keyword-vector. At runtime, we detect the keyword by extracting the same feature vector from a sliding window and computing a simple similarity score between this test vector and the keyword vector. With clean speech, we achieve 86% relative false rejection rate reduction at 0.5% false alarm rate when compared to a competitive phoneme posteriorgram with dynamic time warping KWS system, while the reduction in the presence of babble noise is 67%. Our system has a small memory footprint, low computational cost, and high precision, making it suitable for on-device applications.

show abstract

Continuous hidden Markov modeling for speaker-independent word spotting

Cited by 197 publications

References 2 publications

Low resource point process models for keyword spotting using unsupervised online learning

Low resource point process models for keyword spotting using unsupervised online learning

Feature analysis for discriminative confidence estimation in spoken term detection

Query-by-example keyword spotting using long short-term memory networks

Contact Info

Product

Resources

About