JPH0155477B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention relates to a word speech recognition method for recognizing words by comparing input speech with a word dictionary in which phonemes are expressed.

(Constitution of Conventional Example and Problems thereof) FIG. 1 is a functional block diagram of an apparatus for executing an example of a conventional word speech recognition method and an embodiment of the word speech recognition method of the present invention. Conventional example first
This will be explained with reference to FIG. In Figure 1, 1 is a parameter extraction unit that creates a time series of parameters from input speech, 2 is a probability density calculation unit that calculates the probability density of a phoneme by comparing standard phoneme patterns, and 3 is a segmentation unit for each phoneme. This is a word recognition unit that performs calculations such as probability calculation, likelihood calculation, and word similarity calculation. In addition, 4 is a phoneme standard pattern that stores a phoneme standard pattern that represents the distribution of various parameters for each phoneme in the form of an average value for each phoneme (〓 _i ) and a covariance matrix (Σ _i ) between various parameters. Section 5 is a word dictionary section in which a word dictionary in which all words to be recognized are expressed in symbol strings in units of phonemes is stored. In the word dictionary, for example, the words "Satsuporo" and "Asahikawa" are written as "SAQPORO", "ASAHIKAWA", etc.

FIG. 2 illustrates the movement of the likelihood value of each phoneme that occurs when speech is uttered from the XYZ phoneme sequence.

Next, the operation of the above conventional example will be explained. The parameter extraction unit 1 analyzes input phonemes every 10 ms frame, extracts parameters, and creates a parameter time series. Next, the probability density calculation section 2 compares the parameters obtained for each frame with the phoneme standard pattern of the phoneme standard pattern section 4, and calculates the probability density of the phoneme generated from the parameter values. Next, the word recognition unit 3 performs phoneme segmentation for each dictionary item according to the dictionary phoneme series that makes up the dictionary item, and then segments the phoneme according to the type of phoneme and the phoneme corresponding to the phoneme according to the following formula. The likelihood l of the segmented interval is calculated, and the similarity is determined as the average of the likelihoods of each phoneme in the dictionary entry. Here, the phoneme is
If the frame numbers at the start and end of the segmented section for X are N _s and N _e , and the value of each parameter in the nth frame is C _o , then the likelihood l of phoneme _X is Defined by Eq.

φ _i (C _o ) represents the probability density of a certain phoneme i and is defined as in the equation.

C: n parameters (vector) in one frame 〓 _i : Average parameter value (vector) of a certain phoneme i Σ _i : Covariance matrix In the formula, the range of i in the denominator summension in probability density division is: It varies depending on the phoneme X. For example, when X is the phoneme A (a), the range of i is five vowels, A, E, I, O, and U. The word similarity L _M obtained above was determined for each dictionary item according to the formula, and the dictionary item with the maximum L _M was selected as a recognized word.

L _M = _NP 〓 ⁱ⁼¹ l _i /NP … L _M Similarity of the Mth word in the dictionary l Likelihood of phoneme i in _{the i} dictionary phoneme series NP Number of dictionary phonemes In the above conventional example, segmentation As shown in Figure 2, in the section where phoneme
Other phonemes j ₁ and j ₂ appear with values comparable to the probability density value of X, so when calculating the likelihood of phoneme X according to the formula, only the probability density value of phoneme X is included in the numerator. Since it is not taken into account, sufficient likelihood cannot be obtained,
As a result, this became a factor in word recognition errors.

(Object of the Invention) The present invention is intended to eliminate the drawbacks of the above-mentioned conventional example, and aims to improve the accuracy of likelihood calculation and thereby improve the word recognition rate.

(Structure of the Invention) In order to achieve the above object, the present invention provides the phoneme
When calculating the likelihood of
By performing likelihood calculations that incorporate the probability density values of phonemes other than phoneme X, the accuracy of likelihood calculations can be improved.

(Description of Examples) Examples of the present invention will be described below. The phoneme standard pattern and word dictionary in this embodiment are the same as in the conventional example. Furthermore, the parameter time series obtained by the parameter extraction step is also the same as in the conventional example.

First, a parameter extractor 1 obtains parameters for each frame from the input speech, and then, using the values of the parameters, a probability density calculator 2 calculates the probability density obtained from each phoneme standard pattern. The steps up to this point are the same as in the conventional example. Next, in the word recognition unit 3, segmentation of the phoneme X is performed for each dictionary item in the word dictionary unit 5 according to the dictionary phoneme series that constitutes the dictionary item, and the phoneme X and the phoneme The likelihood _lX of the corresponding segmented interval is calculated. Taking into consideration the probability density of phonemes other than the phoneme X appearing in (φ _i1 , φ _j2 of j ₁ and j ₂ in FIG. 2), the likelihood l _X is determined according to the formula.

N _s , N _e : Starting and ending frame numbers of the segmented section C _o : Value of each parameter in the nth frame φ : Probability density defined by the formula Here, W (X, Y, Z, n) is This is a weighting function determined by the previous phoneme Y, the subsequent phoneme Z in the dictionary phoneme sequence of phoneme X, and the frame position within the segmentation interval.

Further, the range of j in the second term of the numerator Σ ⁱ φ _j (C _o ) is set according to the phoneme appearing in the crossing part. W
The values of (X, Y, Z, n) and j are determined in advance through preliminary experiments and the like. The definition of the denominator i is the same as the expression. As described above, the word similarity L _M is determined for each dictionary item according to the formula in the same way as in the conventional example from the obtained likelihood value, and the dictionary item with the maximum L _M is determined as a recognized word.

In this example, by taking into account the preceding and following phonemes in the dictionary entry within the segmented interval, and performing likelihood calculations using the probability density values of phonemes that appear in the crossing part, It has the advantage of obtaining a highly accurate likelihood.

(Effects of the Invention) The present invention has the above-described configuration, and provides the following effects. Within the segmented interval, by taking into account the preceding and following phonemes in the dictionary entry, and performing a likelihood calculation using the probability density value of the phoneme that appears in the crossing part,
A highly accurate likelihood can be obtained.

[Brief explanation of drawings]

FIG. 1 is a diagram for explaining the word speech recognition method in the conventional method and in an embodiment of the present invention, and FIG. 2 shows the phoneme X including the preceding and following phonemes Y and Z, and the phoneme j ₁ appearing in the intervening part, FIG. 3 is a diagram showing temporal changes in the probability density φ of j ₂ . 1...Parameter extraction unit, 2...Probability density calculation unit,
3... Word recognition section, 4... Phoneme standard pattern section, 5... Word dictionary section.

Claims (1)

[Claims] 1 In a word recognition method that performs word recognition of input speech using a word dictionary that describes the word to be recognized as a symbol string for each phoneme, and a standard pattern for each phoneme that is expressed as a distribution of the parameters of each phoneme. , the input speech is compared with each dictionary entry in the word dictionary, the input speech is segmented for each phoneme according to the dictionary phoneme series that constitutes each dictionary entry, and the segmentation is performed using the standard pattern of that phoneme. When calculating the likelihood for a segment of speech, the value of the probability density of a phoneme that appears in a transitional part within the segmented segment of speech is calculated according to the adjacent phonemes before and after in the dictionary phoneme sequence. A word speech recognition method that calculates the included likelihood and uses this likelihood value to find the degree of similarity between a dictionary entry and input speech to recognize the word.