JP2583854B2 - Voiced / unvoiced judgment method - Google Patents
Voiced / unvoiced judgment methodInfo
- Publication number
- JP2583854B2 JP2583854B2 JP61172043A JP17204386A JP2583854B2 JP 2583854 B2 JP2583854 B2 JP 2583854B2 JP 61172043 A JP61172043 A JP 61172043A JP 17204386 A JP17204386 A JP 17204386A JP 2583854 B2 JP2583854 B2 JP 2583854B2
- Authority
- JP
- Japan
- Prior art keywords
- voiced
- value
- variable
- unvoiced
- threshold value
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Fee Related
Links
Description
【発明の詳細な説明】 産業上の利用分野 本発明は音声合成，音声認識等の音声処理に用いるこ
とのできる、音声波形の有声無声判定方法に関するもの
である。Description: BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a voiced / unvoiced determination method for a voice waveform, which can be used for voice processing such as voice synthesis and voice recognition.
従来の技術 近年、コンピューターの発達と共に、音声合成，音声
認識等の音声処理技術の開発が急速に進められてきてい
る。なかでも、音声の有声無声判定はピッチ周期抽出と
同様に音声処理を行う際には不可欠な技術である。2. Description of the Related Art In recent years, with the development of computers, the development of speech processing techniques such as speech synthesis and speech recognition has been rapidly advanced. Above all, voiced / unvoiced determination of voice is an indispensable technique when performing voice processing in the same manner as pitch period extraction.
以下図面を参照しながら、上述した従来の有声無声判
定の一例について説明する。第６図は従来の有声無声判
定を説明するためのフローチャートを示すものである。Hereinafter, an example of the above-described conventional voiced / unvoiced determination will be described with reference to the drawings. FIG. 6 is a flowchart for explaining the conventional voiced / unvoiced judgment.
まず、変形相関法によって音声信号の自己相関関数値
υτと線形予測係数αi (P)とから、残差信号の自己相関
関数値γτを求める。以下、変形相関法について説明す
る。まず、線形予測係数の相関関数として、式（１）で
示す逆スペクトル係数を求める。次に、この逆スペクト
ル係数と音声信号の自己相関関数値で式（２）に示す計
算を行い、残差信号の自己相関関数値を求める。First, since the modified correlation method and the autocorrelation function value upsilon tau audio signal and the linear prediction coefficient alpha i (P), determining the autocorrelation function value gamma tau of the residual signal. Hereinafter, the deformation correlation method will be described. First, an inverse spectrum coefficient represented by Expression (1) is obtained as a correlation function of the linear prediction coefficient. Next, the calculation shown in equation (2) is performed using the inverse spectrum coefficient and the autocorrelation function value of the audio signal, and the autocorrelation function value of the residual signal is obtained.
ここに、p:分析次数 次に、残差信号の自己相関関数値γτをピッチ周期が
存在すると思われる遅れ時間Tmin〜Tmaxについてすべて
求め、その中で最大であるγτmaxを求める。 Where, p: analysis order Then, I calculated all the delay time Tmin~Tmax you think that the pitch period of the autocorrelation function value gamma tau of the residual signal is present, the maximum and is gamma tau max therein.
有声音か無声音かの判定は、相関関数値γτmaxを残
差信号の零次の自己相関関数値γ０で正規化した値ρma
x（＝γτmax/γ０）と一次のPARCOR係数k1から以下の
式に示す識別関数を求め、この値をもとに経験的に判断
する。The determination of voiced sound or unvoiced sound is made by normalizing the correlation function value γ τ max with the zero-order autocorrelation function value γ 0 of the residual signal ρma
obtains an identification function shown in the following equation from x (= γ τ max / γ 0) and primary PARCOR coefficient k 1, is determined empirically based on this value.
ρmax＋k1/2≧0.4→有声音 ρmax＋k1/2＜0.4→無声音 （中田和男，三瓶 徹共編，「音声デバイス活用の実
際」，オーム社）。ρmax + k1 / 2 ≧ 0.4 → voiced sound ρmax + k1 / 2 <0.4 → unvoiced sound (edited by Kazuo Nakata and Toru Sampei, “Practical Use of Voice Devices”, Ohmsha).
発明が解決しようとする問題点 しかしながら、上記のような方法では、線形予測係数
αi (P)、音声信号の自己相関関数値υτ、残差信号の自
己相関関数値γτおよび、一次のPARCOR係数k1を求める
ことが必要で、これらは多くの乗算を行わなければなら
ず、これらをパソコンを用いて処理しようとすると、非
常に時間がかかるという問題点を有していた。Problems to be Solved by the Invention However, in the above method, the linear prediction coefficient α i (P) , the autocorrelation function value υ τ of the audio signal, the autocorrelation function value γ τ of the residual signal, and the first order is necessary to obtain the PARCOR coefficients k 1, they must perform many multiplications, when these attempts to process using a computer, has a problem that very time consuming.
本発明は上記問題点に鑑み、パソコンを用いても短時
間で処理できて、しかも精度の良い有声無声判定方法を
提供するものである。SUMMARY OF THE INVENTION In view of the above problems, the present invention provides a highly accurate voiced / unvoiced determination method that can be processed in a short time even with a personal computer.
問題点を解決するための手段 上記問題点を解決するために本発明の有声無声判定方
法は、標本化した音声波形の所定区間において、標本値
の絶対値和と標本値の和を求めるだけで、有声無声判定
を行うものである。Means for Solving the Problems In order to solve the above-mentioned problems, the voiced / unvoiced determination method of the present invention employs a method for determining the sum of absolute values of sample values and the sum of sample values in a predetermined section of a sampled speech waveform. And voiced / unvoiced judgment.
作用 本発明は上記した方法によって、主な演算は標本化し
た音声波形の所定区間における標本値の和を求める加算
になるので、処理時間が短くなる。According to the present invention, according to the above-described method, a main operation is an addition for obtaining a sum of sample values in a predetermined section of a sampled audio waveform, so that a processing time is shortened.
実施例 以下本発明の一実施例の有声無声判定方法について、
図面を参照しながら説明する。The following describes a voiced / unvoiced determination method according to an embodiment of the present invention.
This will be described with reference to the drawings.
第１図は本発明の一実施例における有声無声判定方法
を説明するための音声波形図の一例を示すものである。
第１図で区間12の音声波形は有声音の音声波形を示し、
区間13の音声波形は無声音の音声波形を示す。この図か
ら明らかなように、有声音は周期的な波形から成り、無
声音は周期性のない不規則な波形化に成る。さらに、有
声音と無声音とでは明らかに振幅の違いがみられる。有
声音は無声音に比較し、大振幅であり、従って単位時間
当たりのエネルギー（音圧）も大きくなる。本願発明は
このような音声波形の性質を利用するものである。FIG. 1 shows an example of a voice waveform diagram for explaining a voiced / unvoiced determination method according to an embodiment of the present invention.
In FIG. 1, the voice waveform in section 12 shows the voice waveform of a voiced sound.
The voice waveform in the section 13 shows a voice waveform of an unvoiced sound. As is apparent from this figure, the voiced sound has a periodic waveform, and the unvoiced sound has an irregular waveform with no periodicity. Furthermore, there is a clear difference in amplitude between voiced and unvoiced sounds. A voiced sound has a larger amplitude than an unvoiced sound, so that the energy (sound pressure) per unit time is also larger. The present invention utilizes such a property of the audio waveform.
第２図は本発明の一実施例における有声無声判定方法
を説明するためのフローチャートを示すものである。FIG. 2 is a flowchart for explaining a voiced / unvoiced determination method according to an embodiment of the present invention.
本発明の有声無声判定方法について、以下第１図およ
び第２図を用いてその方法を説明する。The voiced / unvoiced determination method of the present invention will be described below with reference to FIGS. 1 and 2.
標本の精度は8bit、標本化周波数は10kHz、判定区間1
1は20msである。ゆえに、１判定区間における標本数は2
00である。有声無声判定は１区間毎に行う。まず、変数
Ａを零に初期化する。また、変数Ａと比較するためのし
きい値を40に初期設定する。変数Ａに判定区間11の始端
における標本値を加え、変数Ａの絶対値をしきい値と比
較する。変数Ａの絶対値がしきい値より大きければ、判
定区間11は有声音であると判定し、判定区間11の有声無
声判定を終了する。もし、変数Ａの絶対値がしきい値と
等しいか小さければ、さらに次の標本値を加えて、変数
Ａの絶対値としきい値とを比較する。変数Ａの絶対値が
しきい値より大きくなるまで、変数Ａに標本値を加える
ことと、変数Ａの絶対値をしきい値と比較することを繰
り返すが、判定区間11の終端に達しても変数Ａの絶対値
がしきい値を超えなければ、判定区間11は無声音である
と判定する。この操作を音声区間の始端から終端まで行
う。第３図は本発明による有声無声判定を用いて、音声
の有声無声判定を行い、有声音と判定された音声区間の
ピッチ周期抽出を行った例を示す図である。Sampling accuracy is 8 bits, sampling frequency is 10 kHz, judgment interval 1
1 is 20 ms. Therefore, the number of samples in one judgment interval is 2
00. The voiced / unvoiced determination is performed for each section. First, the variable A is initialized to zero. Also, a threshold value for comparison with the variable A is initialized to 40. The sample value at the beginning of the determination section 11 is added to the variable A, and the absolute value of the variable A is compared with a threshold value. If the absolute value of the variable A is larger than the threshold value, it is determined that the determination section 11 is a voiced sound, and the voiced / unvoiced determination in the determination section 11 ends. If the absolute value of the variable A is equal to or smaller than the threshold value, the next sample value is further added, and the absolute value of the variable A is compared with the threshold value. Until the absolute value of the variable A becomes larger than the threshold value, adding the sample value to the variable A and comparing the absolute value of the variable A with the threshold value are repeated. If the absolute value of the variable A does not exceed the threshold value, it is determined that the determination section 11 is an unvoiced sound. This operation is performed from the beginning to the end of the voice section. FIG. 3 is a diagram showing an example in which voiced / unvoiced determination of voice is performed using voiced / unvoiced determination according to the present invention, and a pitch period of a voice section determined to be voiced is extracted.
以上のように本実施例によれば、判定区間における、
標本値の加算としきい値との比較だけの演算なので、処
理時間が短く、8bit CPUパソコンでも充分実用に耐える
ことができ、しかも第３図に示すように精度の良い有声
無声判定を行うことできる。As described above, according to the present embodiment,
Since it is only an operation of adding a sample value and comparing it with a threshold value, the processing time is short, and it can withstand practical use even with an 8-bit CPU personal computer. In addition, as shown in FIG. 3, accurate voiced / unvoiced judgment can be performed. .
以下本発明の第２の実施例について図面を参照しなが
ら説明する。Hereinafter, a second embodiment of the present invention will be described with reference to the drawings.
第４図は本発明の第２の実施例における有声無声判定
方法を説明するための音声波形図の一例を示すものであ
る。FIG. 4 shows an example of a voice waveform diagram for explaining a voiced / unvoiced determination method according to the second embodiment of the present invention.
第５図は本発明の第２の実施例における有声無声判定
方法を説明するためのフローチャートを示すものであ
る。FIG. 5 is a flowchart for explaining a voiced / unvoiced determination method according to the second embodiment of the present invention.
本発明の有声無声判定方法について、以下第４図およ
び第５図を用いてその方法を説明する。The voiced / unvoiced determination method of the present invention will be described below with reference to FIGS. 4 and 5.
第１の実施例と同様に標本の精度は8bit、標本化周波
数は10kHz、判定区間21は20msである。ゆえに、１判定
区間における標本数は200である。有声無声判定は１区
間毎に行う。まず、変数Ａを零に初期化する。次に判定
区間21における標本各々の絶対値を求め、標本200個の
絶対値の和を求める。この絶対値和に従って、変数Ａと
比較するためのしきい値を初期設定する。すなわち、絶
対値和が150未満であればしきい値を64、絶対値和が250
以上499未満であればしきい値を40、絶対値和が600以上
3585未満であればしきい値を40、絶対値和が3585以上で
あればしきい値を90と設定し、絶対値和が150以上249未
満、もしくは500以上599未満であれば、判定区間22にお
いて用いたしきい値と同じしきい値を設定する。以下、
第１の実施例と同様に変数Ａに標本値を加算しながら、
変数Ａの絶対値和としきい値とを比較し、判定区間21の
終端に達するまでに変数Ａの絶対値和がしきい値より大
きくなれば、判定区間21を有声音区間、しきい値より大
きくならなければ無声音区間と判定する。この操作を音
声区間の始端から終端まで行なう。As in the first embodiment, the sampling precision is 8 bits, the sampling frequency is 10 kHz, and the determination section 21 is 20 ms. Therefore, the number of samples in one judgment section is 200. The voiced / unvoiced determination is performed for each section. First, the variable A is initialized to zero. Next, the absolute value of each sample in the determination section 21 is obtained, and the sum of the absolute values of 200 samples is obtained. A threshold value for comparison with the variable A is initially set according to the absolute value sum. That is, if the sum of absolute values is less than 150, the threshold is 64, and the sum of absolute values is 250
If the number is less than 499 and less than 499, the threshold is 40, and the sum of absolute values is 600 or more
If the absolute value sum is less than 3585, the threshold value is set to 40; if the absolute value sum is 3585 or more, the threshold value is set to 90; if the absolute value sum is 150 or more to less than 249, or 500 to less than 599, the determination section 22 The same threshold value as that used in is set. Less than,
While adding the sample value to the variable A as in the first embodiment,
The sum of the absolute value of the variable A is compared with the threshold, and if the sum of the absolute values of the variable A becomes larger than the threshold before reaching the end of the determination section 21, the determination section 21 is set to the voiced section and the threshold. If it does not increase, it is determined to be an unvoiced sound section. This operation is performed from the beginning to the end of the voice section.
以上のように本実施例によれば、判定区間における、
標本値の加算としきい値との比較だけの演算なので、処
理時間が短く、8bit CPUパソコンでも充分実用に耐える
ことができ、しかも判定区間における絶対値和に従って
しきい値を設定しているので、音声のレベルが低い判定
区間でも、音声波形に規則性があれば有声音と判定され
ることとなり、精度の良い有声無声判定を行うことがで
きる。As described above, according to the present embodiment,
Since it is only an operation of adding the sample value and comparing with the threshold value, the processing time is short, and it can withstand practical use even with an 8-bit CPU personal computer, and since the threshold value is set according to the absolute value sum in the judgment section, Even in the determination section where the voice level is low, if the voice waveform has regularity, it is determined that the voice is a voiced sound, so that accurate voiced / unvoiced determination can be performed.
なお、第１の実施例においてしきい値は40としたが、
これは標本値のビット精度に従って変える。また、第２
の実施例におけるしきい値も同様に標本値のビット精度
に従って変える。Although the threshold value is set to 40 in the first embodiment,
This varies according to the bit precision of the sample value. Also, the second
Similarly, the threshold value in the embodiment of the present invention is changed according to the bit precision of the sample value.
発明の効果 以上のように本発明は、入力音声波形を所定の周期で
標本化し、あらかじめ零に初期化しておいた変数Ａに、
所定区間の始端の標本値から順次標本値を加えることと
し、上記変数Ａに標本値を加える毎に、上記変数Ａの絶
対値と、あらかじめ定めておいた正のしきい値とを比較
することとし、上記変数Ａの絶対値が上記しきい値より
大きければ、上記所定区間は有声音と判定して、上記変
数Ａに標本値を加えることを終了し、上記変数Ａの絶対
値が上記しきい値と等しいか小さければ、さらに標本値
を加えて上記しきい値と比較する操作を繰り返し、上記
所定区間の終端までの標本値を加えても、上記変数Ａが
上記しきい値を超えなければ、上記所定区間を無声音と
判定するようにしたこと、および、有声無声判定しよう
とする所定区間において標本値の絶対値をすべて加えた
値をあらかじめ算出しておき、上記標本値の絶対値の和
に従って、しきい値を定めるようにしたこと、および、
有声無声判定しようとする所定区間および、上記区間に
先行する区間各々において、標本値の絶対値をすべて加
えた値をあらかじめ算出しておき、上記２区間における
絶対値の和に従って、しきい値を定めるようにしたこと
により、主な演算は、標本化した音声波形の所定区間に
おける標本値の和を求める加算になるので、処理時間が
短くなり、8bit CPUパソコンでも充分実用に耐えること
ができ、しかも判定区間における絶対値和に従ってしき
い値を設定しているので、音声のレベルが低い判定区間
でも、音声波形に規則性があれば有声音と判定されるこ
ととなり、精度の良い有声無声判定を行うことができ
る。As described above, according to the present invention, the input speech waveform is sampled at a predetermined cycle, and the variable A, which has been initialized to zero in advance,
Adding sample values sequentially from the sample value at the beginning of the predetermined section, and comparing the absolute value of the variable A with a predetermined positive threshold value each time a sample value is added to the variable A If the absolute value of the variable A is larger than the threshold value, the predetermined section is determined to be a voiced sound, and the addition of the sample value to the variable A is terminated. If the threshold value is equal to or smaller than the threshold value, the operation of adding the sample value and comparing with the threshold value is repeated, and even if the sample value up to the end of the predetermined section is added, the variable A must not exceed the threshold value. For example, the predetermined section is determined to be unvoiced, and a value obtained by adding all the absolute values of the sample values in the predetermined section to be voiced or unvoiced is calculated in advance, and the absolute value of the sample value is calculated. Threshold according to the sum That we have to define a, and,
In each of the predetermined section to be voiced and unvoiced and the section preceding the above section, a value obtained by adding all the absolute values of the sample values is calculated in advance, and the threshold value is set in accordance with the sum of the absolute values in the above two sections. By doing so, the main operation is the addition that calculates the sum of the sampled values in the predetermined section of the sampled audio waveform, so the processing time is shortened, and even an 8-bit CPU personal computer can withstand practical use, In addition, since the threshold is set according to the sum of absolute values in the determination section, even in the determination section where the voice level is low, if the voice waveform has regularity, it is determined that the voiced sound is present. It can be performed.
第１図は本発明の第１の一実施例における有声無声判定
方法を説明するための音声波形図、第２図は本発明の第
１の一実施例における有声無声判定方法を説明するため
のフローチャート、第３図は本発明による有声無声判定
を用いて、音声の有声無声判定を行い、有声音と判定さ
れた音声区間のピッチ周期抽出を行った例を示す説明
図、第４図は本発明の第２の一実施例における有声無声
判定方法を説明するための音声波形図、第５図は本発明
の第２の一実施例における有声無声判定方法を説明する
ためのフローチャート、第６図は従来の有声無声判定方
法を説明するためのフローチャートである。 11……判定区間、12……有声音部、13……無声音部、21
……判定区間、22……判定区間。FIG. 1 is a voice waveform diagram for explaining a voiced / unvoiced determination method in a first embodiment of the present invention, and FIG. 2 is a voice waveform / voicelessness determination method in a first embodiment of the present invention. FIG. 3 is a flow chart, FIG. 3 is an explanatory diagram showing an example in which voiced / unvoiced determination of voice is performed using the voiced / unvoiced determination according to the present invention, and a pitch period of a voice section determined to be voiced is extracted, and FIG. FIG. 5 is a voice waveform diagram for explaining a voiced / unvoiced determination method according to the second embodiment of the present invention. FIG. 5 is a flowchart for explaining a voiced / unvoiced determination method according to the second embodiment of the present invention. 5 is a flowchart for explaining a conventional voiced / unvoiced determination method. 11: judgment section, 12: voiced part, 13: unvoiced part, 21
... Determination section, 22... Determination section.
Claims (3)
た所定の区間毎に決定する有声無声判定方法であって、
入力音声波形を所定の周期で標本化し、あらかじめ零に
初期化しておいた変数Ａに、所定区間の始端の標本値か
ら順次標本値を加えることとし、上記変数Ａに標本値を
加える毎に、上記変数Ａの絶対値と、あらかじめ定めて
おいた正のしきい値とを比較することとし、上記変数Ａ
の絶対値が上記しきい値より大きければ、上記所定区間
は有声音と判定して、上記変数Ａに標本値を加えること
を終了し、上記変数Ａの絶対値が上記しきい値と等しい
か小さければ、さらに標本値を加えて上記しきい値と比
較する操作を繰り返し、上記所定区間の終端までの標本
値を加えても、上記変数Ａが上記しきい値を超えなけれ
ば、上記所定区間を無声音と判定するようにしたことを
特徴とする有声無声判定方法。1. A voiced / unvoiced determination method for determining voiced / unvoiced of a voice waveform for each predetermined section designated in advance,
The input voice waveform is sampled at a predetermined cycle, and a sample value is sequentially added from a sample value at the beginning of a predetermined section to a variable A that has been initialized to zero in advance. Each time a sample value is added to the variable A, The absolute value of the variable A is compared with a predetermined positive threshold value.
If the absolute value of the variable A is larger than the threshold value, the predetermined section is determined to be a voiced sound, the addition of the sample value to the variable A ends, and whether the absolute value of the variable A is equal to the threshold value is determined. If the variable A does not exceed the threshold value even if the sample value up to the end of the predetermined interval is added even if the sample value up to the end of the predetermined interval is added, the operation is repeated. Is determined as a voiceless sound.
無声判定しようとする所定区間において標本値の絶対値
をすべて加えた値をあらかじめ算出しておき、上記標本
値の絶対値の和に従って、しきい値を定めることとした
特許請求の範囲第（１）項記載の有声無声判定方法。2. A method in which a threshold value is variable, wherein a value obtained by adding all absolute values of sample values in a predetermined section in which voiced / unvoiced determination is to be performed is calculated in advance, and the absolute value of the sample values is calculated. The voiced / unvoiced determination method according to claim 1, wherein the threshold value is determined according to the sum.
無声判定しようとする所定区間および、上記区間に先行
する区間各々において、標本値の絶対値をすべて加えた
値をあらかじめ算出しておき、上記２区間における絶対
値の和に従って、しきい値を定めることとした特許請求
の範囲第（１）項記載の有声無声判定方法。3. A method in which a threshold value is variable, and in each of a predetermined section in which voiced / unvoiced determination is to be performed and a section preceding the above section, a value obtained by adding all absolute values of sample values is calculated in advance. The voiced / unvoiced determination method according to claim 1, wherein the threshold value is determined according to the sum of the absolute values in the two sections.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| JP61172043A JP2583854B2 (en) | 1986-07-22 | 1986-07-22 | Voiced / unvoiced judgment method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| JP61172043A JP2583854B2 (en) | 1986-07-22 | 1986-07-22 | Voiced / unvoiced judgment method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| JPS6327900A JPS6327900A (en) | 1988-02-05 |
| JP2583854B2 true JP2583854B2 (en) | 1997-02-19 |
Family
ID=15934473
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| JP61172043A Expired - Fee Related JP2583854B2 (en) | 1986-07-22 | 1986-07-22 | Voiced / unvoiced judgment method |
Country Status (1)
| Country | Link |
|---|---|
| JP (1) | JP2583854B2 (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2734526B2 (en) * | 1988-06-16 | 1998-03-30 | ヤマハ株式会社 | Pitch extraction device |
-
1986
- 1986-07-22 JP JP61172043A patent/JP2583854B2/en not_active Expired - Fee Related
Also Published As
| Publication number | Publication date |
|---|---|
| JPS6327900A (en) | 1988-02-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Moorer | The optimum comb method of pitch period analysis of continuous digitized speech | |
| US4038503A (en) | Speech recognition apparatus | |
| EP0302663B1 (en) | Low cost speech recognition system and method | |
| US20040133424A1 (en) | Processing speech signals | |
| KR100298300B1 (en) | Method for coding audio waveform by using psola by formant similarity measurement | |
| US8942977B2 (en) | System and method for speech recognition using pitch-synchronous spectral parameters | |
| US6470311B1 (en) | Method and apparatus for determining pitch synchronous frames | |
| JP2779325B2 (en) | Pitch search time reduction method using pre-processing correlation equation in vocoder | |
| KR20060047451A (en) | Method for determining variable length of frame for preprocessing of a speech signal and method and apparatus for preprocessing a speech signal using the same | |
| JP2583854B2 (en) | Voiced / unvoiced judgment method | |
| JP3354252B2 (en) | Voice recognition device | |
| Samad et al. | Pitch detection of speech signals using the cross-correlation technique | |
| US4783808A (en) | Connected word recognition enrollment method | |
| JP2001083978A (en) | Speech recognition device | |
| JPH0246960B2 (en) | ||
| JPS61128300A (en) | Pitch extractor | |
| Sankar | Pitch extraction algorithm for voice recognition applications | |
| US20060149539A1 (en) | Method for separating a sound frame into sinusoidal components and residual noise | |
| JP2643202B2 (en) | Detection device for steady, transient and uncertain parts of input speech | |
| KR100212453B1 (en) | Method for detecting the pitch of voice signal using quantization error | |
| JPS63124100A (en) | Fundamental frequency analyzer | |
| JPS60159798A (en) | Voice recognition equipment | |
| JPS63195700A (en) | Formant extractor | |
| JPS6042799A (en) | Formant peak extractor | |
| JPH041920B2 (en) |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| LAPS | Cancellation because of no payment of annual fees |