Education and demonstration – Language – Speech
Patent
1994-07-06
1997-10-21
Apley, Richard J.
Education and demonstration
Language
Speech
395751, G09B 500
Patent
active
056790015
DESCRIPTION:
BRIEF SUMMARY
This invention concerns a childrens speech training aid. Such an aid is useful in comparing a child's pronunciation with a desired value to give an indication of correct pronunciation. This indication may then be used by an operator, eg a teacher or therapist etc, to identify children in need of speech therapy or by a child to change pronunciation until acceptable.
BACKGROUND
There are a number of speech recognisers of differing complexity and accuracy. Most identify a single word out of a stored set of words, and if that word is identified, activate a further step, eg operate a switch. Many speech recognisers require extensive training by the operator and operate with a few set words. When asked to speak a single word many children make additional sounds such as uumm or er, and vary the time taken to speak. Additionally, in a classroom environment there are various non speech background noises, and possibly a teachers voice reading out words for a non reader. This makes speech recognition more difficult, because the non speech sounds themselves have to be recognised and dismissed as superfluous. Many available speech recognisers will give a match even when only part of a desired word is spoken. Thus for example the word "county" might be recognised as "cow", "count", or "tea", or correctly as "county".
Each word in a spoken language can be broken down into a set of about 40 phonemes, ie 40 different sounds, that can be use alone or combined together to pronounce each word. Some phonemes are context insensitive, others are context sensitive, ie they sound different according to the pre- and proceeding phonemes combinations. The convention adopted to describe these phonemes is described for example by J Wells et al, in a book chapter Specification of SAM phonetic alphabet (SEA), included in: P Winski, W J Barry &A Fourien (Ed), Support available from SAM project for other ESPRIT speech and language work, The SAM Project, Dept of Phonetics, University College, London.
The present invention is concerned with the correctness of a complete spoken word against a desired standard in the midst of non speech or noise sounds.
STATEMENT OF INVENTION
According to the present invention a child's speech training aid compares a child's speech with that speech as generated by stored acoustic models and general non speech sounds to give an indication of whether or not the child has spoken correctly.
According to this invention a child's speech training aid comprises an acoustic model store for storing acoustic models of context sensitive phonemes representing the processed results of speech by numerous children; a dictionary for storing a list of words together with information to construct acoustic models of these words from the acoustic model store; a network generator for generating an acoustic model of a required word together with an acoustic model of, general speech sounds; input means for requesting the generation of an acoustic model of the required word; means for receiving a child's utterances; means for processing the child's utterances into a form suitable for comparison with acoustic models of speech; a speech pattern matcher for comparing the processed child's utterances with the acoustic models of the required word and general speech; and means for indicating whether or not the child's utterances are the required word together with an indication of the accuracy of pronunciation of that required word. The training aid acoustic models may be Hidden Markov Models.
The child's utterances are sampled and processed into an N-plurality of channels to provide an N-dimensional feature vector in successive time slot periods forming a sequence of feature vectors.
The speech pattern matcher may compute the relative probabilities of the sequence of feature vectors being either the same as the acoustic word model or the acoustic general speech model.
The speech pattern matcher may operate continually so that the required word is recognised within continual speech.
The speech pattern matcher may compute the accumulat
REFERENCES:
patent: 3345979 (1967-10-01), Miura et al.
patent: 5027406 (1991-06-01), Roberts et al.
patent: 5202952 (1993-04-01), Gillick et al.
patent: 5212730 (1993-05-01), Wheatley et al.
patent: 5333236 (1994-07-01), Bahl et al.
J.N. Holmes, "Speech Synthesis & Recognition," 1988, pp. 103-168.
ICASSP-92 IEEE International Conference on Acoustics, Speech and Signal Processing, 23 Mar. 1992, IEEE Signal Processing Society, San Francisco, vol. 2, Lorin P. Netsch et al., "Speaker Verification Using Temporal Decorrelation Post-Processing", pp. II-181 to II-184.
J. Vandewalle et al., "Signal Processing VI Theories and Applications Proceedings of EUSIPCO-92," 24 Aug. 1992, Elsevier, Brussels, vol. 1, T. Kuhn et al., Context-Dependent Modelling in a Two-Stage HMM Word Recogniser for Continuous Speech, pp. 439-442.
Luis Torres, "Signal Processing V Theories and Applications Proceedings of EUSIPCO-90," 18 Sep. 1990, Elsevier, Barcelona, Spain, vol. 3, Alberto Ciaramella et al., A Recognizer for Large Vocabulary Applications, pp. 1503-1506.
Radio Fernsehen Elektronik, vol. 40, No. 6, 1991, Berlin DD, pp. 314-318, Rudiger Hoffman, "Neue Wege Der Spracherkennung".
Russell Martin J.
Series Robert W.
Wallace Julie L.
Apley Richard J.
Richman Glenn E.
The Secretary of State for Defence in Her Britannic Majesty's Go
LandOfFree
Children's speech training aid does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Children's speech training aid, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Children's speech training aid will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-1002071