JP2679039B2 - Vowel cutting device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION Industrial field In vowel pronunciation practice using a language training device, a speech training device, etc., not only analysis and evaluation of a vowel alone but also more natural words Analysis and evaluation are desired. The present invention relates to a vowel cutout device for cutting out a vowel analysis frame in a spoken word containing only one vowel as a learner's vowel in such a language practice device or a speech training device. 2. Description of the Related Art Conventionally, processing based on a signal level or processing based on a spectrum change amount is used for vowel cutout. Among them, the one based on the amount of spectrum change is that the LPC cepstrum coefficient or the like is calculated for each frame, and then the coefficient distance is set between adjacent frames, and accurate segmentation can be performed, but a great amount of calculation time is required. Therefore, signal level processing is often used as a method for easily extracting vowels. Hereinafter, an example of the conventional vowel cutout device described above will be described with reference to the drawings. FIG. 2 is a block diagram of essential parts of a conventional vowel cutout device. In the figure, 21 is a level detection unit, and 22 is a level peak detection unit. The operation of the conventional vowel cutout device configured as described above will be described below. It is assumed that the absolute value of the waveform amplitude is used as the parameter indicating the signal level. In the level detection unit 21, the signal level Pn, n = 1 of each frame is sequentially applied from the start frame 1 to the end frame N.
~ N is calculated. here X ₁ ... X _m : the above-mentioned signal level sequence in the voice waveform data level peak detection unit 22 of the nth frame
The maximum frame is searched from P ₁ ... P _{N and} the necessary number of frames before and after is cut out as vowel frames. (See FIG. 3) Problems to be Solved by the Invention In the vowel cutout device having the above-mentioned configuration, even if the frame position where the level is the maximum is different between speakers or the same speaker in the vowel section. It varies greatly with each utterance. The present invention has been made in view of the above points, and an object of the present invention is to provide a vowel cutout device capable of easily cutting out a stable vowel frame. Means for Solving the Problems In order to achieve the above object, the present invention provides a level detection unit that detects a signal level of input word speech waveform data that includes only one single vowel as a vowel for each frame of a fixed time length. A cumulative level calculating unit for calculating a cumulative value of the signal level in each frame from the start frame, and a cumulative level for detecting a frame in which the ratio of the total cumulative value up to the end frame exceeds a threshold value set in advance for each word. And a comparison section. Action The present invention has the above-described configuration and determines the vowel cutout frame based on the ratio of the cumulative value of the signal level. In addition, since the consonant before and after the vowel is determined by the input word name as prior knowledge for the variation of the vowel section for each input word voice, it is possible to respond by changing the ratio depending on the combination of the consonant before and after this vowel. is there. Embodiments A vowel cutting device according to an embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram of essential parts showing an embodiment of a vowel cutout device of the present invention. In FIG. 1, 11 is a level detection unit, 12 is a cumulative level calculation unit, and 13 is a cumulative level comparison unit. The operation of the vowel cutout device configured as described above will be described below. It is assumed that the absolute value of the waveform amplitude is used as the parameter indicating the signal level. Similar to the level detecting unit 21 in the conventional example, the level detecting unit 11 sequentially calculates the signal levels Pn, n = 1 to N of each frame from the starting frame 1 to the ending frame N. The cumulative level calculator 12 calculates cumulative values Qn, n = 1 to N from the start frame of the signal level in each frame. In the cumulative level comparing unit 13, a threshold value Kj, j set for each word: word number and total cumulative value Q _N up to the end frame
The ratio Q _i / Q _N of the cumulative value Q _i up to each frame is compared to detect a frame in which the ratio exceeds a threshold value, and the necessary number of frames before and after are cut out as vowel frames. Fig. 4 shows an example of a threshold of 50%. The threshold Kj is set to the above-mentioned cumulative level ratio in the vowel frame cut out when the vowel cutout is performed by the processing by the spectral change amount for the word voice containing only one single vowel as the vowel of many speakers for each word. This is an average of the above values, and this value has less fluctuation for each speaker and utterance as compared with the conventional method by level peak detection, and a stable vowel frame can be cut out. EFFECTS OF THE INVENTION As described above, according to the present invention, by providing the cumulative level calculating section and the cumulative comparing section, it is possible to easily and stably perform vowel cutout.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a block diagram of essential parts in a vowel cutting device according to an embodiment of the present invention, FIG. 2 is a block diagram of essential parts in a conventional vowel cutting device, and FIG. The graph shown in FIG. 4 is a graph showing the cumulative value of the signal level. 11 …… Level detector, 12 …… Cumulative level calculator, 13 ……
Cumulative level comparison unit, 21 ... Level detection unit, 22 ... Level peak detection unit.

Claims (1)

(57) [Claims] A level detector that detects the signal level of the input word speech waveform data that contains only one vowel as a vowel, for each frame of a fixed time length, and a cumulative value that calculates the cumulative value of the above signal level in each frame from the start frame A vowel cutout device comprising: a level calculation unit; and a cumulative level comparison unit that detects a frame whose ratio to the total cumulative value up to the end frame exceeds a threshold value set in advance for each word.