1271702 九、發明說明: 【發明所屬之技術領域】 本發明係關於-種產生合成聲音之音高模式之技術。 【先前技術】 對應於中國話之聲音合成裝置中,裝設有依輸入之拼音 (±以羅馬字將中國話之讀法拼音化者)而輸出中國話之合成 聲音之功能。 >此日',中國話係!個漢字與1個音節對應,^個音節包含: 稱為聲母」之最前子音(在音節最前之子音),及稱為「韻 :」之除去「聲母」之部分(母音、雙重母音、鼻音化母音 寻)。 為了獲付中國話之合成聲音’需要以羅馬字輸人(拼音輸 :)此種聲母與韻母,不過中國話中存在多數個具有相同拼 二之漢字。如某個音節「qi」,即有「期」、「奇」、「起」、·· · $ ’即使僅輸人拼音,仍無法立即獲得需要之轉換輸出候 補0 ^了解決此種問題’而與拼音合併採用輸入表示音節之 二(:間:之音高變化)之稱為「四聲」之聲調(聲調資訊) 二#音輸人方法(如參照專敎獻丨)。該聲調基本 匕3 .維持其音高(音之高度)之第一聲,提高音高_ 二將音高暫時降低後再度提高之第三聲及降低音高之; 四卑(翏照圖16)。輪入声u田吹> + 弟 )輸耳调負矾時,係將第一聲〜第四聲之 弇凋附加於對應以卜4 作說明,獲得「期」(-第抑。列舉1 』」(―弟一荦)、「奇」(=第二聲)、「起」(= 95459.doc a^17〇2 候補情況下,係分 。如此,#由與拼 成為單一指定對應 第一聲)、「器」(==第四聲)作為轉換輸出 :輪出為「qil」、「qi2」、rqi3」、「qi4」 9合併輸入表示聲調種類之聲調資訊, 於拼B之漢字及意義之線索。 [專利文獻1]特開昭61-27597號公報 【發明内容】 可依輸人之聲調獲得各音節之音高變化,但匡 「;:广周與則後音節聲調之關係(如該音節之聲㈣ ::」,而後續之音節聲調為「第二聲」等),而存名 上述g向變化不自然等之問題。 二卜’除藉由使用者指^聲調之種類,來改變合成聲韦 之…卜’亦需要自由改變合成聲音之音高等。 自亡述之情況’本發明之第一目的在提供一種實頻 供二種音高模式產生技術,其第二目_ 術。 布主之曰円變化用之音高模式產生括 為了解決上述問題,太欢〇口 ^ _ A # 、么月之曰南模式產生裝置之特德 為·係依據輸入之文字資 ^ _ 貝讯,產生表示對應於該文字資郭 S之曰兩之時間性變化之音高模式,且且備:承 得手段,其係自前述文字次1 —立— 八備承 一 貝讯,母曰郎取得表示基準音高 / 定貧訊,及表示聲調種類之聲調資訊;記憶手段, ,、係將聲調編號,標準音高變化模式,及改變該標準音高 變化核式^變形^變化模式相對應而記憶;選擇手段, 八係自取传之音即之聲調資訊指定前述聲調編號,且自該 95459.doc 1271702 音節之前之音節之聲調資訊或後續之音節之聲調資訊,選 擇對應於前述聲調編號之前述標準音高變化模式或前述變 形音高變化模式之任何-個;及產生手段,其係依據選擇 之任何-個音高變化模式與取得之音節之音高指定資訊, 而產生該音節之音高模式。 」采用該構造’係自取得之音節之聲調資訊(如「第 Γ定聲調編號,且自該音節之前之音節之聲調資訊❹ =㈣之音節之聲調資訊’選擇對應於該聲調編號之伊 !;=模•「第三聲」之標準之音高變化模式)或: 二準“變化模式之變形音高變化模式之任何—個 圖8及圖9)。如此’由於係選擇除該音節之聲 亦考慮前後音節之聲調之音高變卜 咅筋夕辣,冰、踢讲* 一 ^ U此與僅考慮該 咖式時比較,可獲得更自然之 树明之音高模式產生裝置之特徵為 二文:貧訊’產生表示對應於該文字資訊之合成聲 …時間性變化之音高模式,且具備::二之 自前述文字資訊,每音節取得表示 係 訊,及表示聲調種類之聲調資訊;却/…兩指定資 編號及標準音高變化模式°思'手段’其係將聲調 式產生手段,其係自取:變形音高變化模 編號’抽出對應於該聲調編號之標前述聲調 由依據該音節之前之音節之聲調資訊之::?’並藉 貧訊來改變抽出之標準音 、曰即之聲調 %式,而產生變形音高變 95459.doc 1271702 Γ:二:二音高模式產生手段,其係依據產生之前述變形 :立::::式與取得之音節之音高指定資訊來產生該音節 人之文卜字_=明:::模式產生裝置之特徵為:係依據輸 音 門 、不對應於該文字資訊之合成聲音之 :::性變化之音高模式’且具傷:取得手 自…字資訊,每音節取得表示基準音高之 : 讯,檢測手段,其係檢測 1曰貝 記憶手段,其係將"記號4:1:;=^ 愫·撰搂车仍. 〜一日阿,交化杈式相對應而記 重立資又^、係就檢測出前述重音資訊之音節,自該 二:;述重音記號,而選擇對應於該重音記號之 及產生手段’其係依據選擇之前述音高變 =式與檢測出前述重音資訊之音節之前述音高資訊,來 產生該音節之音高模式。 採用該構造,就檢測出重音資訊之音節,係自該重音資 戒指定重音記號’而選擇對應於指定之重音記號之音高變 1 匕模式(參照圖11及圖12)。如此,由於選擇反映重音資訊内 =音高變化模式等,因此可獲得模式化之聲調無法表現 之曰咼變化及使用者希望之音高變化。 此外,本發明之音高模式產生裝置之特徵為:係依據輸 入之文字資訊’產生表示對應於該文字資訊之合成聲音之 音兩之時間性變化之音高模式’且具備:第—取得手段, 其係自前述文字資訊’每音節取得表示基準音高之音高指 定貧訊:檢測手段,其係檢測前述各音節中是否包含重立 95459.doc 1271702 取得手段,其係自前述文字資訊 頁訊;第 前述重音資訊之音節’取得表示聲調種::聲:::測出 一記憶手段,其係將重音記號與音高變化模^=’·第 憶,·第二記憶手段,其係將聲調、^應而記 應而記憶;第一選擇手π 曰-交化模式相對 ^ 擇手奴,其係就檢測出前述重音資却 音即,自該重音資訊指定前 …之 會立㈣夕一 料菫曰°己唬,而選擇對應於該 重“己说之音面變化模式;第二選擇手段 :1271702 IX. DESCRIPTION OF THE INVENTION: TECHNICAL FIELD OF THE INVENTION The present invention relates to a technique for generating a pitch mode of synthesized sound. [Prior Art] The sound synthesizing device corresponding to the Chinese language is provided with a function of outputting the synthesized sound of the Chinese language according to the input pinyin (± the pronunciation of the Chinese word in Roman characters). >This day, Chinese language! The Chinese characters correspond to one syllable, and the syllables include: the foremost consonant called the consonant (the first consonant in the syllable), and the part called "the rhyme" that removes the "consonant" (vowel, double vowel, nasalization) Mother sound search). In order to be paid for the synthetic voice of the Chinese language, it is necessary to input the initials and the finals in Roman characters, but there are many Chinese characters with the same spell in Chinese. For example, if a certain syllable is "qi", there are "period", "odd", "start", ··· $ ' even if only the pinyin is lost, the conversion output candidate is not immediately available. ^^ Solve this problem' In combination with Pinyin, the input is used to indicate the syllable of the second syllable (the difference between the pitch and the pitch) (called the tone of the four sounds) (tune information). The tone is basically 匕3. Maintain the first sound of its pitch (the height of the sound), and increase the pitch _ 2. Temporarily lower the pitch and then increase the third sound and lower the pitch; ). When you turn in the sound of the sound of the field, you can add the first sound to the fourth sound, and then add the corresponding sound to the corresponding 4 to give the "period" (- the first suppression. List 1) "("一弟一荦", "奇奇" (= second voice), "起起" (= 95459.doc a^17〇2 in the case of an alternate, the system is divided. Thus, # is the first designation corresponding to the spell. Sound), "器" (== fourth sound) as the conversion output: the round is "qil", "qi2", rqi3", "qi4" 9 combined input tone information indicating the type of tone, in the Chinese characters of B [Patent Document 1] JP-A-61-27597 [Summary of the Invention] The pitch change of each syllable can be obtained according to the tone of the input, but 匡 ";: the relationship between the wide and the subsequent syllables ( For example, the sound of the syllable (4)::", and the subsequent syllable tone is "second sound", etc.), and the name of the g-direction changes unnaturally. To change the synonym of the sound... Bu's also need to freely change the pitch of the synthesized sound, etc. The situation of the death is described in the first purpose of the present invention. A kind of real frequency for two kinds of pitch pattern generation technology, the second item _. The pitch mode used by the cloth master changes to solve the above problems, too happy mouth ^ _ A #, 么月之曰The genre of the south mode generating device is based on the input character _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ From the above-mentioned texts 1 - Li - Ba Bei Cheng Yi Bei, the mother Lang obtains the reference pitch / fixed poor news, and the tone information indicating the type of tone; memory means, ,, the tone number, standard pitch Change mode, and change the standard pitch change kernel type ^ deformation ^ change mode corresponding to the memory; selection means, the eight-series self-received tone, that is, the tone information specifies the aforementioned tone number, and from the 95459.doc 1271702 syllable The tone information of the syllable or the tone information of the subsequent syllable, selecting any one of the aforementioned standard pitch change patterns or the aforementioned pitch pitch change patterns corresponding to the aforementioned tone number; and generating means, which are selected according to Any of the pitch change patterns and the pitch of the obtained syllables specify the information, and the pitch mode of the syllable is generated. "This structure is used as the tone information of the obtained syllables (such as "the first tone number, and The syllable information from the syllable before the syllable ❹ = (4) The syllable information of the syllable 'Select the Iraqi number corresponding to the tone number!; = MODE • The third pitch of the standard pitch change mode) or: Any of the modes of the deformation pitch change mode - Figure 8 and Figure 9). So because of the selection of the sound of the syllable, the pitch of the syllables before and after the syllable is changed. ^ U This is compared with the case of considering only the coffee type, and the more natural tree-like pitch pattern generating device is characterized by two texts: the poor news 'generates the pitch representing the temporal change of the synthesized sound corresponding to the text information. Mode, and has:: two from the above text information, each syllable to obtain a representation of the tone, and tone information indicating the tone type; but / ... two designated capital number and standard pitch change mode ° thinking 'means' its tone Generation Means, the self-fetching: the deformation pitch change mode number 'extracts the tone corresponding to the tone number. The tone is determined by the tone information according to the syllable before the syllable::?' and changes the extracted standard sound by the poor news,曰 之 % % , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , Information to generate the syllable character of the syllable _= Ming::: The mode generating device is characterized in that: according to the sound door, the synthesized sound that does not correspond to the text information::: the pitch mode of the sexual change' Injury: Get the word information from the hand, and obtain the reference pitch for each syllable: News, detection means, which is a means of detecting 1 mussel memory, which will be written by "mark 4:1:;=^ 愫· The car is still. ~ One day, the accommodating 杈 杈 相对 相对 相对 相对 相对 , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , Generating means 'based on the selection of the aforementioned pitch change = formula and detecting the foregoing Syllable tone pitch information of the foregoing information to generate the pitch pattern of syllables. With this configuration, the syllable of the accent information is detected, and the accent mark corresponding to the designated accent mark is selected from the accent tone or the accent mark is selected (see Figs. 11 and 12). In this way, since the selection reflects the accent information = pitch change mode, etc., it is possible to obtain a change in the tone that cannot be expressed by the moded tone and a pitch change desired by the user. In addition, the pitch mode generating device of the present invention is characterized in that: according to the input text information 'generates a pitch pattern indicating a temporal change of the synthesized sound corresponding to the text information, and has: a first means of obtaining , from the above-mentioned text information 'per syllable to obtain the pitch of the reference pitch to specify the poor news: detection means, which is to detect whether the above syllables include the re-establishment 95459.doc 1271702 acquisition means, from the aforementioned text information page The first syllable of the accent information 'acquisition indicates the tone type:: sound::: a memory means is measured, which is a change of accent marks and pitches ^^'················································ The tone, ^ should be remembered and remembered; the first choice hand π 曰 - cross mode relative to ^ choose the slave, the system will detect the above-mentioned accented voice, that is, from the accent information specified before ... the standing (4) On the eve of the evening, the 菫曰 唬 唬 唬 唬 唬 唬 唬 唬 唬 唬 唬 唬 唬 唬 唬 选择 选择 选择 选择 选择 选择 选择
述聲調資訊之音節,自取得 /、糸就取侍河 丁 < θ即之耷凋貧訊指定前诚声史 調編號’而選擇對應於該聲調編號之音高變化模式;二 產生手段’其係依據藉由前述第一選擇手段選擇之音高變 化模式與檢測出前述重音資訊之音節之前述音高資訊,而 =錢即之音高模式;及第二產生手段,其係依據藉由 ^弟^選擇手段選擇之音高變化模式與取得前述聲調資 訊之音節之前述音高資訊,而產生該音節之音高模式。 如以上之說明,依據本發明可實現自然之音高變化或使 琦者希望之音高變化。 【實施方式】 以下 面翏照圖式一面說明關於本發明之實施形態。 Α·本實施形態 圖1係顯示關於本實施形態之對應於中國話之聲音合成 衣置100之力犯構造之圖。本實施形態係假定安裝於行動電 ^ PHS(個人手機系統··登錄商標)及PDA(個人數位助理) 等對硬體貧源限制較大之攜帶式終端機之情況,不過並不 限定於此,亦可適用於各種電子機器。 95459.doc 1271702 輸入部210將自圖上未顯示之操作部等輸入之文字資訊 供給至文字分析部220。圖2及圖3係例示使用附帶四聲之拼 音輸入方法而輸入之文字資訊之圖。 文字資訊大致上區分為:第一類文字資訊(參照圖2)與第 二類文字資訊(參照圖3),各文字資訊中包含指定合成聲音 之音高(如200(Hz)等)之音高指定資訊(省略圖式)等。The syllable of the tonal information, from the acquisition of /, 取 取 河 河 & θ θ θ θ 耷 耷 指定 指定 指定 指定 指定 指定 指定 指定 指定 指定 指定 指定 指定 指定 指定 指定 指定 指定 指定 指定 指定 指定 指定 指定 指定 指定 指定 指定 指定 指定 指定 指定 指定 指定 指定 指定The method according to the pitch change mode selected by the first selection means and the pitch information of the syllable for detecting the accent information, and the pitch mode of the money; and the second generation means are based on ^ Brother ^ selects the pitch change mode selected by the means and the pitch information of the syllable of the aforementioned tone information to generate the pitch mode of the syllable. As explained above, according to the present invention, it is possible to achieve a natural pitch change or a pitch change desired by the Qi. [Embodiment] Hereinafter, embodiments of the present invention will be described with reference to the drawings.本· EMBODIMENT OF THE INVENTION Fig. 1 is a view showing the structure of the vocal composition corresponding to the Chinese-speaking sound-synthesizing garment 100 of the present embodiment. This embodiment is assumed to be installed in a mobile terminal device such as a mobile phone PHS (personal mobile phone system·registered trademark) and a PDA (personal digital assistant), which are limited in terms of hardware lean source, but is not limited thereto. It can also be applied to a variety of electronic machines. 95459.doc 1271702 The input unit 210 supplies the character information input from the operation unit or the like not shown in the figure to the character analysis unit 220. Fig. 2 and Fig. 3 are diagrams showing text information input using a four-phonetic input method. The text information is roughly divided into: the first type of text information (refer to FIG. 2) and the second type of text information (refer to FIG. 3), and each text information includes the pitch of the specified synthesized sound (eg, 200 (Hz), etc.) High specified information (omitted schema), etc.
第一類文字資訊係不包含後述之重音記號之文字資訊, 並包含:在拼音中附加聲調資訊者(以下總稱為「附聲調拼 音資訊」,參照圖2之A),或其中進一步附加長音記號者(以 下總稱為「附聲調•長音拼音資訊」,參照圖2B)等。 如圖2A所不之文字資訊r xianglgang3(=香港)」係包含附 聲凋拼音資「xiangip香)」與「以叫”^港)」之2音節之 文字資訊,圖2B顯示之文字資訊rcha〇1(=^)__ren2(=仁)」 係包含附聲調•長音之拼音資訊「cha〇1(=超)_·」與附聲調 拼音貧訊「ren2(=仁)」之2音節之文字資訊。 另外,長音記號「_」意味著將該長音記號存在之音節(圖 2之B為「chaol」)僅延長特定長度,連續之長音記號數量 愈多’该音節之發音時間愈長。 另外’第二類文字資訊係包含重音資訊之文字資訊。重 音貧訊係在對應之音節上附加抑揚用之資訊,且包含「,」、 「―」等之重音記號’或表示附加於該重音記號之後之抑揚 強度之「3」、「2」等之重音強度(參照圖3)。 如圖3之A所示之文字資訊 ye3(二也)」中附加重音資訊 12 ye3」係在附聲調拼音資訊 「’2」之1個音節之文字資訊, 95459.doc -10- 1271702 圖3之B所示之文字資訊「,3 ai—2·_,4_」,係在附聲調•長 音拼音貧訊「al(=阿)…」中附加重音資訊「,3」、「―2」、 「’4」之文字貧訊(參照圖4)。s夕卜,由於後面將詳細教述 重音資訊,因此,此處省略說明。 文字分析部220分析自輸入部21〇供給之文字資訊,並將 分析結果分別供給至音高產生部23()、聲音訊號產生部 240。詳述之’文字分析部(取得手段、第一取得手段⑽ 自輸入部2H)取得文字資訊時,藉由將該文字資訊分割成各 音節2分析,而取得表示各音節基準之音高(如2〇〇(Hz)等) 之音高指定資訊、表示音韻音韻資訊及表示音大小或音長 度之韻律資訊。而後,文字分析部22()將分割之每音節之文 字資訊供給至文字資訊種類判斷部231,並且將取得之每音 節之音高指定資訊供給至音高模式產生部236,再將取得之 每音節之音韻資訊及韻律資訊供給至聲音訊號產生部“Ο。 文字資訊種類判斷部(檢測手段)2 3丨判斷自文字分析部 220—供給之每音韻之文字資訊係第一類文字資訊或第二類 文字資訊。文字資訊種類判斷部231於該文字資訊中不含重 音資訊情況下,判斷為第一類文字資訊,另一方面於該文 字^訊中包含重音資訊情況下,判斷為第二類文字資訊。 文字資訊種類判斷部231依據該判斷結果,供給第一類文字 貢訊^聲調資訊取得部仙,並且將第二類文字資訊供給 至重:資訊取得部231b。如此,本實施形態⑷個音節中含 有重音貧訊時,不論該音節中是否包含聲調資訊,均以重 音資訊為優先,而依據該重音資訊執行處理,不過以音節 95459.doc 1271702 中包含之重音資訊為優先,或是以聲調資訊為優先,可依 聲音合成裝置100之設計等來適切變更。 聲調資訊取得部(取得手段、第二取得手段)231a自第一類 文字貝訊取得每音節之聲調資訊,並供給至聲調·音高變 化模式產生部234a。 另外,重音資訊取得部231b自第二類文字資訊取得每音 即之重音貧訊,並供給至重音•音高變化模式產生部U仆。 <聲調·音高變化模式產生部234a〉 聲调·音高變化模式產生部234a包含··聲調•音高變化 杈式透擇部(選擇手段)232a及聲調•音高變化模式表(記憔 手段)233a。 〜 士圖5係例^示聲調•音高變化模式表233a之登錄内容之圖。 :凋曰呵變化模式表(記憶手段、第二記憶手段)233a中將 指^各聲調(第-聲〜第四聲)用之聲調編號與音高變化模 弋刀別相對應而登錄。{高變化模式係表示時間性音高之 者,亚包含:表示各聲調之標準之音高變化之標準音 二义:杈式(苓照圖8及圖9所示之實線部分),及改變對應之 私準曰问夂化枳式之變形音高變化模式(參照圖δ及圖9所 示之虛線部分)。 /又形g Ν嘁化模式係依據之前或後續音節之聲調資訊 與該音節之聲含周:欠# 曰〆 、σ 一 凋貝讯之關係而產生之音高變化模式,圖s 所示之變形音高變4 - 门又化杈式表不具有第三聲以外聲調立 後續時之第二與L «V + 曰即 一 —耳之音高變化,圖9所示之變形音高變化模 表不具有弟一聲之声^:丄田> ^ A/- κ. 耳之茸调之音郎在前時之第二聲之音高變化 95459.doc -12- 1271702 (詳細如後述)。另外’以下之說明,係將依據之前之音節之 聲调魏與該音節之聲調資訊之關係而產生之音高變化镇 式稱為在則型變形音高變化模式,將依據後續音節之 調資訊與該音節之聲調資訊之關係而產生之音高變化振 式,稱為後續型變形音高變化模式。 、 圖6係例示登錄於聲調•音高變化模式表灿之各音高總 化模式之構造圖。 疋The first type of text information does not include the text information of the accent marks described later, and includes: those who add tone information to the pinyin (hereinafter referred to as "acoustic pinyin information", refer to FIG. 2A), or further add a long note (hereinafter referred to as "attached tone + long phonetic information", refer to FIG. 2B) and the like. As shown in Figure 2A, the text information r xianglgang3 (=Hong Kong) contains the text information of the 2 syllables with the sounds of "Sympic" and "Calling" (Hong Kong). Figure 2B shows the text information rcha 〇1(=^)__ren2(=仁)” is a two-syllable text containing the phonetic information “cha〇1(=super)_·” with the tone and long tone and the ninth syllable with the tone of the pinyin “ren2 (=ren)” News. In addition, the long note "_" means that the syllable in which the long note is present ("Bol" in Fig. 2) is only extended by a certain length, and the number of consecutive long notes is increased. The longer the pronunciation of the syllable is. In addition, the second type of text information contains text information of accent information. The accented poor news is attached to the corresponding syllables with information for suppressing, and includes accent marks such as "," "", or "3", "2", etc., which are added to the accent strength after the accent mark. Stress intensity (see Figure 3). As shown in Fig. 3A, the text information ye3 (second also) adds accent information 12 ye3" to the text information of a syllable with the tonal information "'2", 95459.doc -10- 1271702 The text information ", 3 ai - 2 · _, 4_" shown in B, is attached with accent information ", 3", "― 2" in the sound-changing and long-sounding pinyin "al (= Ah)..." The text of "4" is poor (see Figure 4). In the following, since the accent information will be described in detail later, the description is omitted here. The character analysis unit 220 analyzes the character information supplied from the input unit 21, and supplies the analysis result to the pitch generation unit 23() and the audio signal generation unit 240, respectively. When the character analysis unit (the acquisition means and the first acquisition means (10) from the input unit 2H) obtains the character information, the character information is divided into the syllables 2 to obtain the pitch indicating the syllable reference (for example). 2〇〇 (Hz), etc.) The pitch designation information, the rhythm information, and the prosody information indicating the size or length of the sound. Then, the character analysis unit 22() supplies the divided text information for each syllable to the character information type determination unit 231, and supplies the obtained pitch information specifying information for each syllable to the pitch pattern generation unit 236, and acquires each of the acquired The phonological information and the prosody information of the syllable are supplied to the audio signal generating unit "Ο. The text information type determining unit (detecting means) 2 3 丨 the text information of the phonological information supplied from the character analyzing unit 220 - the first type of text information or the first The second type of text information. The text information type determining unit 231 determines that the first type of text information is included when the text information does not include the accent information, and determines that it is the second when the text information includes the accent information. The character information type judging unit 231 supplies the first type of text tweet information tune information acquisition unit based on the determination result, and supplies the second type of character information to the weight: information acquisition unit 231b. Thus, the present embodiment (4) When accent stress is included in a syllable, the accent information is prioritized regardless of whether or not the syllable contains tone information, and the accent is based on the accent The processing is performed, but the accent information included in the syllable 95459.doc 1271702 is prioritized, or the tone information is prioritized, and can be appropriately changed according to the design of the voice synthesizing device 100. The tone information acquisition unit (acquisition means, second The acquisition means 231a obtains the tone information for each syllable from the first type of text, and supplies it to the tone/pitch change pattern generation unit 234a. The accent information acquisition unit 231b obtains the accent per tone from the second type of text information. The poor signal is supplied to the accent/pitch change pattern generation unit U. The tone/pitch change pattern generation unit 234a> the tone/pitch change pattern generation unit 234a includes the tone and the pitch change. Selection (selection means) 232a and tone/pitch change mode table (recording means) 233a. ~ Figure 5 shows the picture of the registration of the pitch/pitch change mode table 233a. In the table (memory means, second memory means) 233a, the tone number for each tone (the first to fourth sounds) is registered in correspondence with the pitch change mode tool. {High change mode The person who expresses the temporal pitch, the sub-inclusion: the standard sound meaning of the pitch change of the standard of each tone: the 杈 type (see the solid line part shown in Figure 8 and Figure 9), and the change of the corresponding private standard曰 夂 变形 变形 变形 变形 ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( : 欠 曰〆, σ 凋 凋 讯 讯 凋 凋 凋 凋 凋 凋 凋 凋 凋 凋 凋 凋 音 音 音 音 音 音 音 音 音 音 音 音 音 音 音 音 音 音 音 音 音 音 音 音 音 音 音 音 音 音2 and L «V + 曰 曰 — — — — 耳 耳 耳 — — — — 耳 耳 耳 耳 耳 耳 耳 耳 耳 耳 耳 耳 耳 耳 耳 变形 变形 变形 变形 变形 变形 变形 变形 变形 变形 变形 变形 变形 变形 变形 变形 变形 变形 变形 变形 变形 变形The pitch of the second sound of the sound of the sound is 95459.doc -12- 1271702 (details are described later). In addition, the following description will be based on the relationship between the tone of the previous syllable and the tone information of the syllable. The pitch change is called the morphological change mode, which will be based on the subsequent syllable information. The pitch-changing mode produced by the relationship with the tone information of the syllable is called a subsequent-type deformation pitch change mode. Fig. 6 is a structural diagram showing the pitch-accumulation mode registered in the tone/pitch change mode table.疋
音高變化模式包含:將賦予音高變化之時間分割成η個時 之各時間U〜tn’及對應於此等之各音高變化量Μ,。另 卜圖6中係例不將賦予音高變化之時間作⑻㈣)等分, 此時之各時間tl=〇 · · ,t31-30,· · ·,ti〇1 = 1〇〇 對應於此等之久立古變几〇 寻又谷曰同、交化置ρ1==1〇,· · ·,ρ31_1〇,· · · ρ101=30 。 ’ 圖7係例不直線插入圖6所示之各時間之各音高變化量等 :獲得之音高變化模式之圖。從圖6及圖7可知,本實施形 態係將賦予音高變化之時間予以等分,來表現上述時間f =不論賦予音高變化之時間的伸縮,均可賦予同樣之音 :夂,。另外’上述例係例示將賦予音高變化之時間予以 等分割之情況’不過並非限定於等分割之意思,只要可藉 At、&直線插人等而獲得音高變化模式,亦可為任何分割 樣。此外’纟高變化模式亦可為@定者,亦可為使用者 自由定義•變更者。 /議示第三聲之音高變化模式之圖,圖9係例示第二 擘之音高變化模式之圖。 95459.doc -13- 1271702 第一聲之標準音高變化模式,表示音高一時降低後再度 提咼之變化(參照圖8所示之實線部分),另外,第三聲之後 績型變形音高變化模式,表示音高降低後維持之變化(參照 圖8所不之虛線部分)。藉由設計該第三聲之後續型變形音 2變化模式,即使在第三聲之音節之後,具有其他聲調: 音節繼續時,仍可獲得自然之音高變化。 聲調·音高變化模式選擇部(選擇手段、第二選擇手 = )232a,自聲調資訊取得部231&取得該音節之聲調資訊 b ’自該聲調資訊指定聲調編號。聲調•音高變化模式選 擇部232a判斷指定之聲調編號係「第三聲」時,參照其後 績之音節之聲調資訊,來判斷後續之音節是否為具有「第 三聲」之聲調之音節。聲調•音高變化模式選擇部2仏依 據該判斷結果’選擇第三聲之標準音高變化模式或第三聲 之後續型變形音高變化模式之任何一個。 如就音節「wu3(=五)及「xiangl gang3(=香港)」中之立 =「卿此港)」,藉由聲調•音高變化模式選擇部‘ 延擇弟三聲之標準音高變化模式,另外,就「μ _g2(= =1U3(,」’及「bei3jingl(=北京)」中 之曰即bei3(=北)」,藉由聲調·立古嶽 、阳摇蝥一 女 曰阿、夂化核式選擇部232a 延擇弟三聲之後績型變形音高變化模式。 另外,第二聲之標準音高變化模式如圖9所示, 高自低位置P職高之變化之模式(“Κ9所示W = 为),而第二聲之在前型變形音高 A、 置PS0高位置之PS1提高之變化 。自比位 之枳式(參照圖9所示之虛線 95459.doc -14- 1271702 部分)。藉由設計該第-款> ^ 蚀丄 一耳之在珂型變形音高變化模式,即 .. 之卓调之音節在前時,藉由自比通常 (亦即具有第一聲之磬 ^ 周之曰郎不在前時)高之位置開始變 化,仍可獲得自然之音高變化。 ^外*亦可亚非母聲調設計在前型變形音高變化模式或 L r里r形音向變化模式之任何一個(參照圖8及圖9),而每 每调設計在前型蠻飛立合h 形曰回受化模式及後續型變形音高變化 兩者。此外,參照聲調資訊之音節並不限定於如上述 之雨-個或後一個音節’亦可為前兩個及後六個音節等。 此外’亦可參照適切組合此等之數個音節之各聲調資訊。 聲調·音高變化模式選擇部(選擇手段、第二選擇手 ^ )232a自聲調資訊取得部23⑽得該音節之聲調資訊 時’自該聲調資訊指定聲調編號。聲調•音高變化模式選 ㈣232a判斷指定之聲調編號為「第二聲」時,參昭在盆 之前音節之聲調資訊,判斷之前之音節是否為具有「第一 聲」之聲調之音節。聲調•音高變化模式選擇部232a依據 該判斷結果,來選擇第二聲之標準音高變化模式或第二聲 之在前型變形音高變化模式之任何一個。 如就「lu3 xing2(=旅行)」中之音節「xing2(=行)」,及 「nei4_g2㈣容)」中之音節「_糾=容)」,藉由聲調· 音高變化模式選擇部232a選擇第二聲之標準音高變化模 式,另外就「anl quan2(=安全)」中之音節「叫奶2(=全)」, 及「zhongl wen2(=中文)」中之音節「we2(=文)」,聲調· 音高變化模式選擇部232a選擇第二聲之在前型變形:高°變 95459.doc -15- 1271702 化模式。The pitch change mode includes each time U to tn' at which the time at which the pitch change is given is divided into n, and the pitch change amount Μ corresponding thereto. In addition, in the example of Fig. 6, the time for imparting the pitch change is not equally divided into (8) (four)), and at this time, each time tl = 〇 · ·, t31-30, · · ·, ti〇1 = 1〇〇 corresponds to this. Wait for a long time to change the ancient times to find a few valleys and the same, the intersection of ρ1 = = 1 〇, · · ·, ρ31_1〇, · · · ρ101=30. Fig. 7 is a diagram in which the pitch change amount and the like of each time shown in Fig. 6 are not linearly inserted: the obtained pitch change pattern. As can be seen from Fig. 6 and Fig. 7, in the present embodiment, the time at which the pitch change is given is equally divided to express the time f = the same sound can be imparted regardless of the time when the pitch is changed. In addition, the above-described example exemplifies a case where the time at which the pitch change is given is equally divided. However, the present invention is not limited to the meaning of equal division, and any pitch change pattern can be obtained by using At, & straight line insertion or the like. Split the sample. In addition, the 'high-change mode can also be set to @, and the user can be freely defined and changed. / A diagram showing the pitch change pattern of the third sound, and Fig. 9 is a diagram illustrating the pitch change pattern of the second sound. 95459.doc -13- 1271702 The standard pitch change mode of the first sound, indicating that the pitch is lowered again after the pitch is lowered (refer to the solid line part shown in Fig. 8), and the third sound after the deformation sound The high change mode indicates the change in the sustain after the pitch is lowered (refer to the dotted line portion of Fig. 8). By designing the subsequent mode of the third sound distortion mode, even after the third sound syllable, there are other tones: When the syllable continues, a natural pitch change can be obtained. The tone/pitch change mode selection unit (selection means, second selection hand = ) 232a, the tone information acquisition unit 231 & acquires the tone information b ’ of the syllable from the tone information. When the tone/pitch change mode selection unit 232a determines that the designated tone number is "third sound", it refers to the tone information of the syllable of the subsequent performance to determine whether the subsequent syllable is a syllable having the "third sound" tone. The tone/pitch change mode selection unit 2 selects any one of the standard pitch change mode of the third sound or the subsequent modified pitch change mode of the third sound based on the determination result '. For the syllables "wu3 (=5) and "xiangl gang3 (=Hong Kong)" = "Qing Hong Kong", by the tone • pitch change mode selection department' Mode, in addition, "μ _g2 (= =1U3 (, "' and "bei3jingl (= Beijing)" is the bei3 (= North)", by tone, Li Guyue, Yang shake a woman夂 核 核 选择 232 232 232 232 232 232 232 232 232 232 232 232 232 232 232 232 232 232 232 232 232 232 232 232 232 232 232 232 232 232 232 232 232 232 232 232 232 232 232 232 232 232 232 232 232 232 ("W9 is shown as )"), and the second sound is in the front-type deformation pitch A, and the change in PS1 at the high position of PS0 is increased. The self-alignment formula (see the dotted line 95459.doc shown in Figure 9) -14- 1271702 Part). By designing the first paragraph > 丄 丄 丄 丄 丄 丄 丄 丄 丄 丄 丄 丄 丄 , , , , , , , , , , , , , , , , , 变形 变形 变形 变形 变形 变形 变形That is to say, when the first sound is 磬 ^ 周之曰郎 is not in the front) the high position begins to change, and the natural pitch change can still be obtained. ^External* can also be Asian and African maternal design Any one of the front-type deformation pitch change mode or the r-shaped r-shaped sound direction change mode (refer to FIG. 8 and FIG. 9), and each of the adjustment designs is in the front type and the fly-shaped h-shaped return mode and the subsequent type. In addition, the syllables of the reference tone information are not limited to the rain-one or the next syllable as described above, and may be the first two and the last six syllables, etc. The tone and pitch change mode selection unit (selection means, second selection hand ^) 232a, when the tone information acquisition unit 23 (10) obtains the tone information of the syllable, 'specifies the tone number from the tone information Tone • Pitch change mode selection (4) 232a When the specified tone number is “Second Sound”, refer to the tone information of the syllable before the basin to determine whether the previous syllable is a syllable with the “first sound” tone. The pitch change mode selection unit 232a selects any one of the standard pitch variation mode of the second sound or the preceding deformation pitch variation pattern of the second sound based on the determination result. For example, "lu3 xing2 (= In the syllable "xing2 (= line)" in the line), and the syllable "_correction" in the "nei4_g2 (four) capacity)", the standard pitch of the second sound is selected by the tone/pitch change mode selection unit 232a Change mode, in addition to the syllable "cream 2 (= all)" in "anl quan2 (= security)", and the syllable "we2 (= text)" in "zhongl wen2 (= Chinese)", tone and pitch The change mode selection portion 232a selects the front type deformation of the second sound: the high degree change 95459.doc -15 - 1271702 mode.
J 另外,就3亥音卽之聲调為「第一聲」時及為「第四聲 時之動作,可與上述大致同樣地說明,因此省略。 聲調·音高變化模式選擇部232a自聲調·音高變化模式 表233a選擇適合聲調資訊之音高變化模式時,將其供給至 音南模式產生部236。 <重音·音高變化模式產生部234b〉 重音·音高變化模式產生部234b包含:重音·音高變化 模式選擇部232b及重音•音高變化模式表23讣。 圖⑺係例示重音•音高變化模式表233b之登錄内容之圖。 在重音•音高變化模式表(記憶手段、第—記憶手段_ ’將重音記號與音高變化模式分別相對應而登錄。圖" 係例示重音記號「,」之音高變 立寸哚「 , 交化杈式之圖,圖12係例示重 田舌己唬「_」之音高變化模式之圖。 如圖η及圖墙示’藉由重音記號 化模式係表*音高逐漸提高㈣ =之曰以 咕「 欠亿之杈式,另外,重音記 唬-」之音高變化模式係声+立a .s,k 式係表不音尚逐漸降低而變化之模 式。另外,就此等音高變化槿 夂 ^ 所示之直绫, 、式,如函數資訊(如為圖11等 厅不之直線% ’為表示斜度及切 登錄於重音•立古銳π 、 、β )寻,只須預先 化模二… 匕模式表233b中即可。另外,音高變 化松式當然亚不限定於直線性者。 a-又 重曰·音尚變化模式選擇部(選擇丰― 段)232b自重音資訊取得部 、擇手&、弟-選擇手 資訊指定登錄於重音·#重音資訊時,自該重音 曰雨受化模式表233b中之重音記 95459.doc 1271702 號,而選擇對應於該重音記號之音高變化模式。而後 音·音高變化模式選擇部232b按照重音資訊所示之重音強 度,變更音高變化模式φ _ 、式中所不之音尚變化量(為圖丨丨及圖i 2 所示之音高變化模式時,#亩 糸直線之斜度),亚依賦予音高 化之時間來變更時間W細内容參照以下說明)。 圖13係例示輸入「,3 ! 2 ^ t ^ al—2—4-」之1個音節之文字資訊(表 照圖3之B等)時之音高變化模式之圖。另外,圖U例示為了 方便說明,而將賦予音高變化之時間設為_時之音高變化 模式。 如圖13所示,賦予音高變化之時間依「ai」、「_」、「_」、 」而作4等刀,並藉由附加於「ai」之重音資訊「,3」而 獲得音高變化ch卜繼續藉由附加於第一個及第三個長二記 號「-」之重音資訊「_2」及「,4」而獲得各個音高變化咖 ch4。不過,由於第二個長音記號「。巾未附加重音資訊: 因此成為音南維持一定值之音高變化ch3。 重音•音高變化模式選擇部2321)如此自重音•音高變化 模式表233b選擇•變更適合重音資訊之音高變化模式時, 將其供給至音高模式產生部236。 ^ 音高模式產生部(產生手段、第—產生手段、第二產生手 段)236依據自聲調•音高變化模式產生部23蝕或重音•音In addition, the sound of the sound of the 3rd sound is "first sound" and the motion of the fourth sound is similar to the above, and therefore the description is omitted. The tone and pitch change mode selection unit 232a is self-tuned. When the pitch change mode table 233a selects the pitch change mode suitable for the tone information, it is supplied to the sound South mode generation unit 236. <Accent/Pitch Change Pattern Generation Unit 234b> Accent/Pitch Change Pattern Generation Unit 234b The accent/pitch change mode selection unit 232b and the accent/pitch change mode table 23A are included. Fig. 7 is a diagram showing the registration contents of the accent/pitch change mode table 233b. In the accent/pitch change mode table (memory) Means, first-memory means _ 'Register the accent mark and the pitch change mode respectively. The figure " is an example of the accent mark "," the pitch of the pitch is changed to "," the map of the cross-cut, Figure 12 The figure shows the pattern of the pitch change pattern of the "_" of the torrent of the tongue. As shown in Figure η and the wall, the pitch is gradually increased by the accent pattern. (4) = after the 咕In addition, the accent record -" The pitch change mode is the sound + vertical a.s, k type is a mode in which the sound is gradually reduced and changes. In addition, the pitch changes as shown by the pitch, 式, such as function information (such as For the line of Figure 11, etc., the % ' is the slope and the cut is registered in the accent • Li Gurui π, , β), only need to pre-module the second... 匕 mode table 233b. In addition, the pitch change The loose type of course is not limited to the linear one. a- and the heavy 曰 音 音 变化 模式 模式 选择 选择 选择 选择 选择 232 232 232 232 232 232 232 232 232 232 232 232 232 232 232 232 232 232 232 232 232 232 232 232 232 232 232 232 232 In the case of #重音信息, the accent note 95459.doc 1271702 in the accented rain mode table 233b is selected, and the pitch change mode corresponding to the accent mark is selected. The post-pitch change mode selection unit 232b follows The accent intensity indicated by the accent information changes the pitch change mode φ _ and the amount of change in the pitch (in the case of the pitch change mode shown in Fig. 2 and Fig. 2), the slope of the line of #亩糸), Yayi gives the time of pitching to change the time. Under instructions). Fig. 13 is a view showing a pitch change pattern when character information (indicated as B of Fig. 3, etc.) of one syllable of ", 3 ! 2 ^ t ^ al - 2 - 4" is input. Further, Fig. U exemplifies a pitch change mode in which the time at which the pitch change is given is _ for convenience of explanation. As shown in Fig. 13, the time for giving the pitch change is 4 for the "ai", "_", "_", and "," and the sound is obtained by adding the accent information ", 3" attached to "ai". The high-change ch-b continues to obtain the individual pitch change ch4 by the accent information "_2" and ", 4" attached to the first and third long-length marks "-". However, since the second long note "the towel is not attached with accent information: it becomes a pitch change ch3 in which the sound is maintained at a certain value. The accent/pitch change mode selection portion 2321] is thus selected from the accent/pitch change mode table 233b. • When the pitch change mode suitable for accent information is changed, it is supplied to the pitch mode generation unit 236. ^ The pitch mode generation unit (generation means, first generation means, second generation means) 236 is based on the tone and pitch Change pattern generation section 23 eclipse or accent
高變化模式產生部234b輸出之音高變化模式,及抽S自I 字分析部220供給之音高變化模式之音節之音高指定資 訊,藉由在基準之指定音高中附加音高變化模式,而產生 如圖14所示之音高模式。 95459.doc -17- 1271702 ^ s ·Λ號產生部240依據自音高模式產生部236供給之音 高模式與自文字分析部220供給之音韻資訊及韻律資訊,而 產生合成聲音訊號。因而,依據如上述產生之音高模式之 *- 合成聲音經由揚聲器(省略圖式)等而輸出至外部。 - 如以上之說明,本實施形態之聲音合成裝置選擇除該音 節之聲調外,還考慮前後音節之聲調之音高變化模式 此,與僅考慮該音節之聲調來選擇音高變化模式時比較, I 可獲得顯示更自然之音高變化之合成聲音。 此外,輸入之文字資訊中含有重音資訊情況下,產生顯 不於該重音貧訊之重音記號及反映重音強度之音高變化模 式。藉此,可獲得顯示模式化之聲調無法表現之音高變化 及使用者希望之音高變化之合成聲音。 . Β.變形例 <變形例1〉 上述本實施形態係說明將各音節之聲調分類成具有四種 • 特徵性音高變化之「四聲」之情況,不過,中國話(普通話) 之音節的聲調中亦存在不具確定之音高變化而輕微發音之 稱為「輕聲」者。此等輕聲如僅藉由不附加聲調資訊之拼 音來標記(「xie4xie(=謝謝)」等),該輕聲亦可仍然維持之 前音節之音高變化模式。另外,本實施形態係假定中國話, 不過亦可適用於泰語及越南語等具有聲調之所有語言。此 . 外,上述本實施形態係說明藉由拼音來輸入文字資^之情 況,不過亦可藉由漢字來輸入文字資訊。此時聲調與本^ 施形態同樣地,亦可使用聲調資訊等來輸入,此外,亦^ 95459.doc -18- 1271702 預先準備將各漢字與聲調相對應之漢字•聲調表等,藉由 參照該漢字·聲調表來指定輸入之漢字之聲調。 <變形例2> 圖15係顯示變形例2之聲調•音高變化模式產生部234a’ 之構造圖。聲調•音高變化模式產生部234a’包含··變形音 高變化模式產生部(產生手段)232a,及聲調•音高變化模式 表(記憶手段)233a,。 與圖5所示之聲調•音高變化模式表233a不同之處在於, 聲調•音高變化模式表233a,中,將指定各聲調(第一聲〜第 四聲)用之聲調編號與表示各聲調之標準之音高變化之標 準音高變化模式相對應而登錄,而不將變形音高變化模式 相對應而登錄。 另外’變形音高變化模式產生部(產生手段)232a,,藉由 改、交自聲调•音高變化模式表233 a,抽出之標準音高變化模 式’而產生變形音高變化模式(參照圖8及圖9之虛線部分)。 詳細而言,變形音高變化模式產生部232a,首先依據自聲調 貧訊取得部231a供給之聲調資訊來指定聲調編號。而後, k形音兩變化模式產生部232a,自聲調•音高變化模式表 233a’抽出對應於指定之聲調編號之標準音高變化模式。 變形音高變化模式產生部232a,抽出標準音高變化模式 牯苓知、5亥音郎之前之音節之聲調資訊(或後續之音節之聲 調貧訊),來決定是否產生變形音高變化模式。另外在作該 決疋呀,只須預先參照登錄產生變形音高變化模式時之原 則(又形原則)之記憶體等來決定即可。變形音高變化模式產 95459.doc 19 1271702 生部2咖進行須產生變形音高變化模式之決定時,參照健 ,於記憶體(省略圖式)等中之變形原則,來適切改變標準音 回變化模式。如此,變形音高變化模式產生部232a,產生圖8 •及圖9等顯不之’k形音高變化模式,並將其供給至音高模式 ,產,^ 236 $外,變形音高變化模式產生部232aj生變形 曰回夂化杈式後之動作’可與本實施形態同樣地說明,因 此省略說明。 • <變形例3> 此外,以上說明之聲音合成裝置i⑽之各功能,係藉由 CPU(或DSP)執行儲存於R〇M等之記憶體中之程式來實 現’因此該程式可記錄於CD_R0M等記錄媒體中分發,亦 可經由網際網路等之通訊網路來分發。 • 【圖式簡單說明】 圖1係顯示本實施形態之聲音合成裝置之功能構造之區 塊圖。 • 圖2係例示使用本實施形態之附帶四聲之拼音輸入方法 而輸入之文字資訊之圖。 圖3係例示使用本實施形態之附帶四聲之拼音輸入方法 而輸入之文字資訊之圖。 ' 圖4係例示本實施形態之重音資訊賦予前後之文字資訊 之圖。 .· 目5係例示本實施形態之聲調•音高變化模式表之登錚 容之圖。 1 圖6係顯示本實施形態之音高變化模式之構造圖。 95459.d〇( -20- 1271702 圖7係例示本實施形態之音高變化模式之圖。 圖8係例示本貫施形悲之第三聲之音高變化模式之圖。 圖9係例示本實施形態之第二聲之音高變化模式之圖。 圖10係例示本貫施形怨之重音•音高變化模式表之圖。 圖Π係例示本實施形態之重音記號之音高變化模式之 圖。 圖12係例示本實施形怨之重音記號之音高變化模式之 圖。 圖13係例示本實施形態之重音記號行之音高變化模式之 圖。 圖14係例示本實施形態之音高模式之圖。 圖15係例示變形例2之聲調•音高模式產生部之構造圖。 圖16係例示中國話之各聲調之音高變化模式之圖。 【主要元件符號說明】 100 聲音合成裝置 210 輸入部 220 文字分析部 230 音高產生部 231 文字資訊種類判斷部 231a 聲調資訊取得部 231b 重音資訊取得部 232a 聲調·音高變化模式選擇部 232af 變形音高變化模式產生部 232b t音·音高變化模式選擇部 95459.doc 21 1271702 233a、 233b 234a \ 234b 236 240 233a’ 聲調•音高變化模式表 重音•音高變化模式表 聲調·音高變化模式產生部 重音·音高變化模式產生部 音高模式產生部 聲音訊號產生部 95459.doc -22-The pitch change pattern outputted by the high change pattern generation unit 234b, and the pitch designation information of the syllable of the pitch change mode supplied from the I-characteristic analysis unit 220, by adding the pitch change mode to the designated pitch of the reference, The pitch mode as shown in Fig. 14 is produced. 95459.doc -17- 1271702 ^ s The apostrophe generating unit 240 generates a synthesized audio signal based on the pitch mode supplied from the pitch mode generating unit 236 and the phoneme information and prosody information supplied from the character analyzing unit 220. Therefore, the synthesized sound according to the pitch mode generated as described above is output to the outside via a speaker (omitted pattern) or the like. - as described above, the sound synthesizing device of the present embodiment selects the pitch change pattern of the pitch of the preceding and lower syllables in addition to the tone of the syllable, and compares it with the tone change mode in which only the pitch of the syllable is considered. I Get a synthetic sound that shows a more natural pitch change. In addition, in the case where the input text information contains accent information, an accent mark that does not show the stress of the accent and a pitch change pattern that reflects the intensity of the accent are generated. Thereby, it is possible to obtain a synthesized sound in which the pitch of the mode can not be expressed and the pitch of the user's desired pitch changes. MODIFICATION MODIFICATION <Modification 1> The above-described embodiment describes the case where the syllables of each syllable are classified into four sounds having four characteristic pitch changes, but the syllables of the Chinese (Mandarin) are used. There are also those in the tone that are called "soft" when they are not pronounced with a certain pitch change. These soft voices are only marked by the pinyin without the tone information ("xie4xie (=thank you)", etc.), and the soft voice can still maintain the pitch change mode of the previous syllable. In addition, this embodiment assumes Chinese, but it can also be applied to all languages having a tone such as Thai and Vietnamese. In addition, the above embodiment describes the case where the character is input by pinyin, but the character information can also be input by the Chinese character. At this time, the tone can be input using the tone information or the like in the same manner as the present embodiment. In addition, it is also prepared in advance to prepare a Chinese character and a tone table corresponding to each of the Chinese characters and the tones by reference. The Chinese character tone table specifies the tone of the input Chinese character. <Modification 2> Fig. 15 is a structural diagram showing the tone/pitch change pattern generation unit 234a' of the second modification. The tone/pitch change pattern generation unit 234a' includes a distortion pitch change pattern generation unit (generation means) 232a and a tone/pitch change pattern table (memory means) 233a. The tone/pitch change pattern table 233a shown in FIG. 5 is different in the tone/pitch change pattern table 233a, and the tone numbers and the respective tone numbers (first to fourth sounds) for each tone are specified. The standard pitch change mode of the standard pitch change of the tone is registered correspondingly, and is not registered corresponding to the modified pitch change mode. Further, the 'deformation pitch change pattern generation unit (generation means) 232a generates a deformation pitch change pattern by changing and passing from the tone/pitch change pattern table 233a and extracting the standard pitch change pattern ' (refer to Figure 8 and Figure 9 are the dotted lines). Specifically, the transformed pitch change pattern generation unit 232a first specifies the tone number based on the tone information supplied from the tone-of-mouth acquisition unit 231a. Then, the k-shaped two-change pattern generation unit 232a extracts a standard pitch change pattern corresponding to the designated tone number from the tone/pitch change pattern table 233a'. The transformed pitch change pattern generating unit 232a extracts the standard pitch change pattern 牯苓, the tone information of the syllable before the 5 hai lang (or the subsequent syllable tone) to determine whether or not the morphing change mode is generated. In addition, in order to make this decision, it is only necessary to refer to the memory of the principle (the shape principle) when the deformation pitch change mode is registered in advance. Deformation pitch change mode production 95459.doc 19 1271702 When the Ministry of Health 2 determines the deformation pitch change mode, refer to the deformation principle in the memory (omitted pattern) to change the standard tone. Change mode. In this manner, the transformed pitch change pattern generating unit 232a generates the 'k-shaped pitch change pattern shown in Fig. 8 and Fig. 9 and supplies it to the pitch mode, and produces a change of the pitch pitch. The operation of the mode generation unit 232aj, the deformation and the subsequent operation, can be described in the same manner as in the present embodiment, and thus the description thereof is omitted. <Modification 3> Further, the functions of the above-described voice synthesizing device i (10) are realized by a CPU (or DSP) executing a program stored in a memory such as R〇M or the like, so that the program can be recorded in It is distributed on recording media such as CD_R0M, and can also be distributed via a communication network such as the Internet. BRIEF DESCRIPTION OF THE DRAWINGS Fig. 1 is a block diagram showing the functional structure of a sound synthesizing device of the present embodiment. • Fig. 2 is a view showing the text information input using the four-phonetic pinyin input method of the present embodiment. Fig. 3 is a view showing the text information input using the four-phonetic pinyin input method of the embodiment. Fig. 4 is a view showing the text information before and after the accent information is given in the embodiment. Fig. 5 shows a diagram of the register of the tone and pitch change mode table of the present embodiment. Fig. 6 is a structural diagram showing a pitch change mode of the embodiment. 95459.d〇( -20- 1271702 Fig. 7 is a diagram illustrating a pitch change pattern of the present embodiment. Fig. 8 is a diagram illustrating a pitch variation pattern of the third sound of the present embodiment. Fig. 10 is a diagram showing a pattern of accent and pitch change patterns of the second embodiment of the present invention. Fig. 10 is a diagram showing the pitch change pattern of the accent marks of the embodiment. Fig. 12 is a view showing a pitch change pattern of the accent mark of the present embodiment. Fig. 13 is a view showing a pitch change pattern of the accent mark line of the embodiment. Fig. 14 is a view showing the pitch of the embodiment. Fig. 15 is a view showing a structure of a tone modulation/pitch mode generation unit according to a modification 2. Fig. 16 is a diagram showing a pitch change pattern of each tone of the Chinese language. [Description of main component symbols] 100 sound synthesis device 210 input unit 220 character analysis unit 230 pitch generation unit 231 character information type determination unit 231a tone information acquisition unit 231b accent information acquisition unit 232a tone/pitch change mode selection unit 232af deformation pitch change mode production Part 232b t-tone pitch change mode selection unit 95459.doc 21 1271702 233a, 233b 234a \ 234b 236 240 233a' Tone • Pitch change mode table accent • Pitch change mode table tone • Pitch change mode generation unit accent · Pitch change mode generation section pitch mode generation section audio signal generation section 95459.doc -22-