JP2514984B2 - Voice recognition system - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Object of the Invention] (Field of Industrial Application) The present invention relates to a speech recognition method capable of obtaining high recognition performance with a small number of learning patterns.

(Prior Art) Input / output of information by voice is highly natural to humans and is excellent as a man-machine interface, and has been variously studied so far. Most of the voice recognition devices that have been put into practical use at present are of the type that recognizes word voices, and are generally configured as shown in FIG.

In this device, the voice input is converted into an electric signal, captured, and acoustically analyzed by an acoustic analysis unit 1 including a band pass filter and the like, and a start / end detection unit 2 detects the word voice section. To do. Then, the pattern matching unit calculates the similarity and distance between the acoustic analysis data (feature information; voice pattern) of the word voice section of the input voice and each standard pattern of the recognition target word registered in the standard pattern dictionary 3 in advance. 4, the determination result is determined by the determination unit 5, and for example, the category name of the standard pattern having the highest similarity value is obtained as the recognition result for the input voice.

However, as described above, in the voice recognition by the pattern matching method, a shift (pattern deformation) in the time axis direction between the input voice pattern and the standard pattern registered in advance becomes a problem. Therefore, in the past, the above-mentioned problem with respect to the shift in the time axis direction has been solved by linear expansion and contraction, nonlinear expansion and contraction represented by dynamic programming (DP), and the like.

On the other hand, apart from such a pattern matching method,
Create an orthogonalization dictionary from the learning patterns collected in advance,
A method of recognizing speech (subspace method) using this orthogonalization dictionary has been proposed. In this method, as shown in the configuration example in FIG. 3, a predetermined number of sample points obtained by equally dividing the voice section by the sample point extraction unit 6 are extracted from the voice pattern subjected to acoustic analysis and voice section detection. , (The number of feature vectors × the number of sample points) is obtained. A predetermined number of such sample patterns are collected for each category to be recognized and stored in the pattern storage unit 7. Then, the Gram-Schmidt (GS) orthogonalization unit 8 creates an orthogonalization dictionary 9 by the following procedure using a predetermined number (three or more) of sample patterns collected in the pattern storage unit 7.

That is, the creation of the orthogonalization dictionary 9 is such that, for each category, the m-th learning pattern of the category is a _m, and when the learning pattern uttered three times is used, the first learning data a ₁ is A one-axis dictionary b ₁ is set and b ₁ = a ₁ (1) This is registered in the orthogonalization dictionary 9.

Using the Gram-Schmidt orthogonalization formula from the _second learning data a ₂ , Then, when ‖b ₂ ‖ is larger than a certain value, this is registered in the orthogonal dictionary 9 as the second axis dictionary b ₂ .
However, (•) is the dot product, T is the transpose, and ‖ ‖ is the norm.

And from the third learning data a ₃ , Then, when ‖b ₃ ‖ is larger than a certain value, this is registered in the orthogonal dictionary 9 as the third axis dictionary b ₃ .
However, when the dictionary of the second axis is not obtained, the calculation of the above formula (2) is performed.

The above processes (1) to (3) are repeatedly executed for each category to form the orthogonalization dictionary 9 in advance.

The similarity calculation unit 10 is provided between the orthogonalization dictionary 9 created as described above and the input voice pattern X. , The similarity between the category i and the orthogonalization dictionary b _{i, r} is calculated, and the input speech pattern X is recognized according to the similarity value. Note that the orthogonal dictionary b _{i, r} of the category i is normalized in advance, and K _i is the category i.
Shows the number of dictionaries (number of axes).

However, in such a method using the GS orthogonalization, there is a problem in that the amount of pattern variation carried by each orthogonal axis is not clear. Therefore, the sample pattern {a _{i, 1} , of the category i of the orthogonalization dictionary 9 calculated as described above is obtained.
There is a problem in that it cannot be guaranteed that a _{i, 2} , a _{i, 3} } expresses the original standard pattern of the category i well.

(Problems to be Solved by the Invention) In the conventional speech recognition by the subspace method using the GS orthogonalization, the orthogonalized dictionary itself includes, for example, the time axis direction of the collected learning patterns and There is a problem caused by fluctuations in the frequency width direction, and a problem remains in terms of whether or not the standard pattern is well expressed. Further, solving such a problem has a drawback that it is necessary to collect a considerably large amount of learning patterns.

The present invention has been made in consideration of such circumstances.
The purpose is to provide a speech recognition method that can improve the recognition performance by creating an orthogonalization dictionary that can sufficiently cope with pattern fluctuations that well expresses the standard pattern with few learning patterns. To do.

[Structure of the Invention] (Means for Solving Problems) The present invention relates to an input speech pattern obtained by analyzing input speech and an orthogonalization dictionary created based on learning patterns collected in advance. In the voice recognition method for recognizing the input voice by calculating the degree of similarity by using three or more types of filters that perform at least smoothing processing and differentiation processing on the learning patterns collected in advance, for example, the collected learning patterns Is obtained, the average pattern is smoothed in the time axis direction and the frequency axis direction to obtain the dictionary of the first axis, and the average pattern is differentiated in the time axis direction to obtain the dictionary of the second axis. The orthogonalization dictionary is created by differentiating the average pattern in the frequency axis direction to obtain a dictionary of the third axis.

(Operation) The average pattern of the collected learning patterns is obtained by using three or more kinds of filters, and the average pattern is smoothed in the time axis direction and the frequency axis direction to obtain the dictionary of the first axis. Axial fluctuations and frequency axial fluctuations can be effectively absorbed. Furthermore, since the average pattern is differentiated in the time axis direction to obtain the dictionary of the second axis, it is possible to effectively absorb the positional deviation of the voice pattern with respect to the time axis direction, and the average pattern is differentiated in the frequency axis direction. Then, since the dictionary of the third axis is obtained, it is possible to effectively absorb the positional deviation of the voice pattern in the frequency axis direction.

In this way, since the orthogonalization dictionary that absorbs the pattern fluctuations in the time axis direction and the frequency axis direction is created, each dictionary pattern of the orthogonalization dictionary can correspond to the positional deviation due to the fluctuation. , Greatly contributes to the improvement of recognition performance. Moreover, the dictionary pattern (first axis) generated from the average pattern that absorbs the pattern fluctuations in the time axis direction and the frequency axis direction is used as the second
Since the orthogonalization dictionary is generated by obtaining the dictionaries for the first axis and the third axis, the pattern variation amount that each orthogonal axis of the orthogonalization dictionary itself does not become unclear as in the conventional art, and there are few learning patterns. It becomes possible to effectively create a high-performance orthogonalization dictionary by effectively using.

Embodiment An embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 is a schematic configuration diagram of a speech recognition apparatus configured by applying an embodiment system according to the present invention. The same parts as those of the conventional apparatus shown in FIG. 3 are designated by the same reference numerals. .

The device of this embodiment is characterized in that, as a means for creating the orthogonalization dictionary 9 using the learning patterns accumulated in the pattern accumulating unit 7, at least the smoothing process and the differential process are performed in place of the conventional GS orthogonalizing unit 8. The point is that the orthogonalization time / frequency filter unit 11 composed of three or more kinds of orthogonalization time / frequency filters for executing Here, as the learning patterns collected in the pattern storage unit 7, for example, j
18 sample points consisting of 16 acoustically analyzed feature vectors represented by (= 1,2, ~ 16) and dividing the speech section into 17 equal parts by k (= 0,1,2, ~ 17) The explanation will be given assuming that it is given as a data series collected over time.

Then, the orthogonalization time / frequency filter unit 11 makes the orthogonalization dictionary 9 as follows when the m-th learning pattern collected for each category i is set to a _{m (j, k)} for the category i. Are being created.

First, the average pattern A _{(j, k)} is calculated from the learning pattern a _{m (j, k) of} category i. Ask as.

After that, the average pattern A obtained as described above
_{Using (j, k)} , b _{1 (j, k)} = A _{(j-1, k-1)} + A _{(j-1, k)} + A _{(j-1, k + 1)} + A _{(j, k) -1)} ＋ 2 * A _{(j, k)} ＋ A _{(j, k + 1)} ＋ A _{(j + 1, k-1)} ＋ A _{(j + 1, k)} ＋ A _{(j + 1, k + 1)} [j ＝ 1,2, to 14, k = 1,2, to 16] (6) The first axis dictionary b _{1 (j, k)} is obtained by the following operation, and this is registered in the orthogonalization dictionary 9. This dictionary b _{1 (j, k)} is obtained by smoothing the average pattern A _{(j, k)} in the time-axis direction and the frequency-axis direction, and is obtained as a reference of the orthogonalization dictionary 9 of the first axis. Registered as dictionary data.

Then, using the average pattern A _{(j, k)} , b _{2 (j, k)} = -A _{(j-1, k-1)} + A _{(j-1, k + 1)} + {-A _{(j , k-1)} + A _{(j, k + 1)} } + {-A _{(j + 1, k-1)} + A _{(j + 1, k + 1)} } [j = 1,2, to 14, k = 1, 2 to 16] (7) The second axis dictionary b _{2 (j, k)} is obtained by the following operation, and this is normalized and then registered in the orthogonal dictionary 9. This dictionary b
_{2 (j, k)} is obtained by differentiating the average pattern A _{(j, k)} in the time axis direction.

The second axis dictionary b _{2 (j, k)} calculated in this way
Is not completely orthogonal to the dictionary b _{1 (j, k) of} the first axis, B _{2 (j, k)} = b _{2 (j, k)} − (b _{2 (j, k)} · b _{1 (j, k)} ) b _{1 (j, k) is} re-orthogonalized, and the re-orthogonalized dictionary data B _{2 (j, k)} is added to the new second axis. The dictionary b _{2 (j, k)} may be registered in the orthogonalization dictionary 9. However, even if such re-orthogonalization is not performed, it is possible to obtain sufficient recognition performance with the dictionary b _{2 (j, k) of} the second axis obtained as described above.

Further, using the average pattern A _{(j, k)} , b _{3 (j, k)} = -A _{(j-1, k-1)} -A _{(j-1, k)} -A _{(j-1, k + 1)} + A _{(j + 1, k-1)} + A _{(j + 1, k)} + A _{(j + 1, k + 1)} [j = 1,2, ~ 14, k = 1,2, ~ 16] ... (8) The third axis dictionary b _{3 (j, k)} is obtained by the following operation, and this is normalized and then registered in the orthogonal dictionary 9. This dictionary b
_{3 (j, k)} is obtained by differentiating the average pattern A _{(j, k)} in the frequency axis direction.

The orthogonalization dictionary 9 is created by repeatedly performing the above-described processes (1) to (5) for each category.

In the above description, an example in which up to three axes are obtained as the orthogonal dictionary 9 has been shown, but it is possible to obtain a 4th order by further performing a second derivative.
You may make it create the dictionary after an axis. In this case, instead of the 18 points described above as the learning pattern, for example,
It suffices to use a sampled point of 20 points or more.

According to the present apparatus that creates the orthogonalization dictionary 9 in this way and recognizes the input voice pattern, the orthogonalization dictionary 9 is used.
Absorbs the fluctuations of the voice pattern in the time axis direction and the frequency axis direction, enabling voice recognition without being affected by the fluctuations of the input voice pattern in the time axis direction and the frequency axis direction. , It is possible to improve the recognition performance. Further, since the orthogonalization dictionary 9 is created by using the orthogonalization time / frequency filter, it becomes possible to efficiently construct an orthogonalization dictionary having high performance with a small number of learning patterns, which has a great practical effect. .

In this way, the present method for creating an orthogonalization dictionary and performing speech recognition by using the differential filter that compensates for the positional deviation in the time axis direction and the frequency axis direction and the orthogonalization filter that absorbs the fluctuation of the two-dimensional pattern According to this, it is possible to effectively compensate for the deviations in the time axis direction and the frequency axis direction, and obtain high recognition performance with a small number of learning patterns. Therefore, it can be said that this method has a great effect on improving the voice recognition performance.

The present invention is not limited to the above embodiment. Here, an example of creating a three-axis orthogonalization dictionary has been described, but it is also possible to create an orthogonalization dictionary with a larger number of axes. In this case, some variations can be considered as the coefficient of the orthogonalization time / frequency filter, but the point is to smooth the learning pattern in the time axis direction and the frequency axis direction, first derivative, second derivative, ... Therefore, various modifications can be implemented. Also, the number of dimensions of the learning pattern is not particularly limited, and the present invention can be modified and implemented without departing from the scope of the invention.

EFFECTS OF THE INVENTION As described above, according to the present invention, an orthogonalization dictionary that absorbs pattern fluctuations in the time axis direction and pattern fluctuations in the frequency axis direction is created using three or more types of filters. Thus, it is possible to obtain a dictionary that effectively expresses the variation of the pattern, and it is possible to achieve a great practical effect such that the recognition performance can be improved.

[Brief description of drawings]

FIG. 1 is a schematic configuration diagram of a voice recognition device configured by applying an embodiment system of the present invention, and FIGS. 2 and 3 are diagrams showing a schematic configuration of a conventional voice recognition device, respectively. 1 ... Acoustic analysis unit, 2 ... Start / end detection unit, 5 ... Judgment unit,
6 ... Sample point extracting unit, 7 ... Pattern accumulating unit, 9 ... Orthogonalizing dictionary, 10 ... Similarity calculating unit, 11 ... Orthogonalizing time / frequency filter.

Claims (2)

(57) [Claims] 1. The input voice pattern is recognized by analyzing the input voice pattern obtained by analyzing the input voice and a degree of similarity between an orthogonalization dictionary created based on a learning pattern collected in advance. In the voice recognition method, the orthogonalization dictionary is created by using three or more types of filters that perform at least smoothing processing in the time axis direction and smoothing processing in the frequency axis direction on a learned pattern and differential processing. Voice recognition method. 2. The filter obtains an average pattern of the collected learning patterns, and smoothes the average pattern in the time axis direction and the frequency axis direction to obtain a dictionary of the first axis, and the average pattern is obtained as a function of time. The method according to claim 1, further comprising means for differentiating in the axial direction to obtain a dictionary for the second axis, and means for differentiating the average pattern in the frequency axis direction to obtain a dictionary for the third axis. Voice recognition method.