EP1261964A1

EP1261964A1 - Method for the voice-controlled initiation of actions by means of a limited circle of users, whereby said actions can be carried out in an appliance

Info

Publication number: EP1261964A1
Application number: EP01921173A
Authority: EP
Inventors: Roland Aubauer
Original assignee: Siemens AG
Current assignee: Siemens AG
Priority date: 2000-03-08
Filing date: 2001-03-08
Publication date: 2002-12-04
Also published as: WO2001067435A9; US20030040915A1; CN1416560A; WO2001067435A1; CN1217314C; DE10011178A1

Abstract

The aim of the invention is to control initiation of actions in a user-independent manner and by means of voice and users pertaining to a limited circle of users of an appliance, whereby said actions can be carried out in the appliance. The voice is detected on the basis of a speaker-dependent voice detection system in a user-independent manner and without user identification. The reference voice patterns of all users pertaining to a voice detection system are allocated to detection voice expressions, e.g. the words of a vocabulary, of the users pertaining to the circle of users, whereby said patterns are required for detection.

Description

description

Method for voice-controlled initiation of actions that can be carried out in a device by a limited group of users

The entry of information or data or commands into a device - e.g. a telecommunications terminal such as a corded or cordless telephone, a cell phone, etc., a household appliance such as a washing machine, an electric cooker, a refrigerator etc., a vehicle such as a car, the airplane, the ship, etc., a consumer electronics device such as the television, the HIFI system, etc., an electronic device for control and command input such as the personal computer, the personal digital assistant, etc. - ^' by means of language, the The natural form of communication of the human being, for the voice-controlled initiation of actions which can be carried out in the respective device, has the primary aim of freeing the hands used for data or command input for other routine activities.

For this purpose, the device has a speech recognition device, which is also referred to in the specialist literature as speech recognizer. The field of automatic recognition of speech as a system of characters and sounds includes recognizing the characters and sounds spoken in isolation - e.g. single words, commands - up to the recognition of fluently spoken characters and sounds - e.g. several connected words, one or more sentences, a speech - according to the form of communication of the person. The automatic speech recognition is in principle a search process, which according to the document

"Funkschau issue 26, 1998, pages 72 to 74" roughly divide into a phase for processing the speech signal, a phase for reducing the amount of data, a classification phase, a phase for forming word strings and a grammar model phase, which in the order given in Speech recognition process running. According to the publication "Funkschau Issue 13, 19998, pages 18 to 80", the speech recognizers working according to this procedure are differentiated according to the degree of their speaker dependency. While in the case of speaker-dependent speech recognizers, the respective user generates the entire vocabulary in at least one learning or training phase speaking of reference patterns, this process does not apply to speaker-independent speech recognizers.

The speaker-independent speech recognizer works almost exclusively on the basis of phonemes, while the speaker-dependent speech recognizer is more or less a single word recognizer.

According to this speech recognizer definition, the speaker-independent speech recognizers are used in particular in devices where, on the one hand, fluently spoken language - e.g. several connected words, sentences etc. - and large to very large vocabulary - i.e. An * unlimited user group uses the device - must be processed and on the other hand the computing and storage effort for recognizing this language and vocabulary does not matter because the corresponding capacities are available.

The speaker-dependent speech recognizers, on the other hand, have their preferred field of use in devices where on the one hand discretely spoken language, eg individual words and B, 1 misses, and small to medium-sized vocabulary - ie a / limited user group uses the device - and on the other hand the arithmetic - And storage effort for the recognition of this language and vocabulary is important because the corresponding capacities are not available. The speaker-dependent speech recognizers are therefore characterized by a low level of complexity in terms of computation and memory requirements. With speech-dependent speech recognizers currently in use, sufficiently high word recognition rates for small to medium-sized vocabularies (10-100 words) are already achieved, so that these speech recognizers are particularly suitable for control and command input (com and-and-control) but also for voice-controlled database access ( eg voice dialing from a phone book). These speech recognizers are therefore being used to an increasing extent in devices on the mass market, such as, for example, in telephones, household appliances, consumer electronics devices, devices with control and command input toys, but also in motor vehicles.

However, the problem with these applications is that the devices are often not only used by one user, but by several users, e.g. frequently members of a household or a family are used (limited number of users).

The problem of the user dependency of speaker-dependent speech recognizers is only solved according to the publication "ntz (news-technical journal) volume 37, issue 8, 1984, pages 496 to 499, in particular page 498, middle column last seven lines to right column first six lines" The disadvantage of the bypass method mentioned is that the users have to identify themselves before using the speech recognition. Since speaker-dependent speech recognition has been assumed, the speaker identification must be carried out using a method other than the speech recognition. The self-identification of the user Using a keyboard and a display is a good option in most cases.Access to automatic speech recognition is made considerably more difficult for the user in terms of user guidance and the time required for speech recognition, especially when there are several e Alternate users of voice recognition frequently. The process of manual user identification even questions the usefulness of speech recognition, since instead of manual user identification with the same effort, the desired execution of the action in the device can be initiated manually and without speech recognition.

The object on which the invention is based is to control the initiation of actions which can be carried out in a device by means of speech by users of a limited group of users of the device, the speech being recognized independently of the user and without user identification on the basis of a speaker-dependent speech recognition system.

This object is solved by the features of claim 1.

The idea on which the invention is based is that the recognition speech utterances of the users of the user group, for example the words of a vocabulary, are assigned the reference speech patterns of all users of a speech recognition system which are necessary for the recognition. The vocabulary (phone book, command word list, ...) contains, for example, "i * words (names, commands, ...), each of which has an action to be performed (phone numbers to be selected, action of a connected device, ...) a possible acoustically outputted voice confirmation (usually the pronunciation of the word) (Voiee Prompt) and up to "j reference speech ^patterns are assigned to the" k ^λλ users of the speech recognition system, where "i * eN," j ^w eN and "k eN ,

The assignment of a voice confirmation to the words of a vocabulary is not absolutely necessary, but is often advantageous for acoustic user guidance. The voice confirmation can come from one of the users of the speech recognition system, a text-to-speech transcription system or, if the words of the vocabulary are certain, from a third person. ω ω)) P ¹ P>

Ül o Cn o cn o Cn

d: T3 <. P »d cn φ N« 3 ö Φ d Q, cn cn S5 rt CΛ S! CΛ i-i Hi vP rt rt PJ W P3 ö

Ö ^J H od Φ rr P- Φ Φ H- Φ H- d Φ Φ g Φ rt 13 JH Ω O 13 φ Φ PJ P- Φ P ¹ P- ι-i P-

Φ O? Φ Hi 3 Φ d H d Φ h d d H d ι-i Φ ι-i Ω PJ tr i-i H d H d φ cn H φ J φ

H 3 PJ H X 3 φ cn Φ iQ N rt s: P J tr P- d rt φ N Φ d P- Ω Φ l-i P-

13 e cn P ⁾ iQ cn f CΛ <! Ω cn cn H- H- P »Ω d P- Φ Ω Cn d Ω tr tr d tr tr rr d P- d Φ s: PJ 13 O tf cn Ω uQ H- d O tr α P- rt 13 NP ⁾ tr Sl Φ P- P-

P- Ω α 3 φ Φ d ι-f li H- 51 13 tr cn rt tr Φ φ φ rt tr Φ ι- <cn fr Φ φ P- O: P- d cn a P> tr Φ H cn d P ⁾ rt- φ O ι-i rt Φ d O d! -f li ii 3 P- HP ⁾ 13 N h Φ HQ

P ⁾ Q n rt 3 P- ι-i Φ Ω Φ ii P ⁾ Φ H- d 3 Φ ri ^¬ Φ d Ω H Φ S! ^ ri ^¬ P- N d Φ H Φ d Φ ded ^J H- Φ rt Ω rt d iQ PJ Φ t? tt) d tr P> 13 o J cn tt) cn d cn CΛ <P- tr ω P- rt Φ Φ l- ¹ d Φ d "Φ rt Φ d ι-id LQ Φ P- p Ω rt li d 13 H rt φ

13 o Ω li P> Ω P- ¹ H- li φ Φ n PP α P- P- d N Φ d cn d tr "H- rt d pj: cn

Φ Φ H tr Φ 3 tr P- f l-i d Hi φ cn d Φ Φ S! d rt n p Φ ^ rt rt Φ LJ.

P- P- tr> H φ rr Ω Φ α cn J »H Ω Φ H d o m • rt d ι-i Φ φ cn cn

13 Ω P> Φ Φ d • tr Φ d Φ s P ⁾ rt Φ tr cn ιQ li Φ Φ cn rt cn s: H cn. 13

Φ tr P P- H ι-id CΛ o tr φ Hi n φ <i - C rt Φ P- α H ri ^¬ • « _* o Φ d -i ϊö

Φ r + <J S! Φ d l-i Φ H cn Φ 13 d O d Φ 1-1 P- d tt) H <P- PJ φ

0- H Φ <o φ P- ω d <! rt H H H l-i ι 3 Φ H s: £ P- O o Ω Ω Hi

Φ rr d Hl O f P- d 13 «3 φ Φ pj: Φ PJ 2! t o-- ω ι-3 Φ Φ H- Φ d fr tr tr Φ d Φ d = ^• P ⁾ rf Hi HH dt P ⁾ d Ω O φ d α P- φ d <Φ pj: d H d PJ rt Φ - (rt £ li PJ tr Φ P »tu Hl H- P- d Cn N tr HH d Φ d P- uq o cn t ü Φ tr H Φ

P ^J Φ s tr d H Ω Ω α J cn Ω l- ¹ cn Φ rt P- φ HP ^J φ dd cn P- dd ^ d

P- H od Φ iT d ^* P>! -R rt tr H- 13 H Φ «5 d • <id Si h- ¹ α O d l- ¹ d φ N

Ω P- H P- P ⁾ Φ Φ cn li • rt Ω o H f ^ ι-f Φ * ⁾ o O P- P ⁾ PJ H φ P ⁾ tr d cn tr Φ rr Φ P ⁾ H <HH Φ tr d ⁾ Φ ^ PJ Φ α Φ ι-i Ω cn cn l- ¹ 3 H Φ d 13

Φ PHO Φ f P ^J PP ⁾ : Ω d φ Cd P ¹ Hi ⁾ φ d ritr φ cn φ ii dt Φ Hi H Φ Φ cn t N _f dd φ Φ tr cn tt) <! PJ d P- P- li P ⁾ ⁾ 3 d φ d = rr D d H- dd Φ ddd φ d Φ cn N o H • d rt d cn Ω

Φ dn P Φ H Φ Φ d Go h ^J H- ¹ P dadd Φ P ^J Q P- dd PJ ι-i iQ 'rf cn dd P- d rr H- H- P ω KQ d rt Cd d ⁾ Φ d p- P3 P- ⁾ rt p cn d cn φ rr r + P> P- ¹ H- d Φ Ω Φ cn rt • ιQ N Φ NH n NO α d H P- dd

? " _^ p- NH) P ^J Φ φ ^ Q t tr 13 φ Φ d cn Φ P ⁾ Φ d Φ Φ P- PJ P- dd ii cn n- P φ PJ-- d Φ HH 11 s: s: H d 13 3 d 3 d P- α P- O Ω rt

P- Φ φ li P ¹ Φ α d Φ H- d N Φ> P- P- rt α rt d «qdd φ rt tr Φ

O ii I- ¹ Φ dd Φ d H Ω N H- li NJ P- <Φ Φ tr CΛ P- Φ ns: H a in φ r - CΛ cn ιQ rr H- Φ tr d Φ Φ O Φ O dd Φ 13 O- φ P ^J Φ Φ.

PJ O ii 13 Hi H dd φ iQ d ι-i ^ li H Φ H P- d H- Φ d H li d H <P ⁾ Φ li Φ SU rt § <j PJ öd <i Φ PJ cn Φ φ * »Α Φ P-

Φ tr PJ φ d Φ Φ Φ - ¹ Φ Q. d ϊö i ^ ¹ o H- d P- C o tr φ o P- Ω dd P- d Λ rr Ω H- α Ω H cn d φ Φ φ H Φ Hi lh pι cn ddd H rt dn iQ 3 Φ tr Ω P ^J H) Φ r ii ϊö Hi d P- cn rt d tr cn Φ 1 3 öd O "* cn cn P- tr r Φ Φ P> CΛ d H- Φ Φ dd φ N Φ 3 J rt Φ H rt φ φ d td 3 J Φ r + cn H & 13 • Hi σ Hi H φ Φ d P- P- cn H φ HN ι-i QK Φ tr d P . φ S!

Φ cn cn. • H H H- Φ Φ H rt cn d φ 13 tr cn Φ φ Φ d -i d φ PJ d o

P- rr x Φ P ⁾ K fr d H d cn φ 1-1 ι-i _—, H _^ ι-i rt cn d φ rt 1-1 ω d li o pi: σ H- α d Ω H- ⁾ d Φ N Φ s: φ d P ⁾ H Φ d ι-i 13 d CΛ HN d cn rt rt tT Φ r P> H Φ cn d "Φ rr rt d cn ι- (Φ ii JÖ d Ω li HP > φ Φ 13 φ Φ N Φ φ H P- α H Φ H H- NN 13 NH d = iQ tr J Φ o rt P- P- ι-i H d PJ Φ cn

1-1 Hl ι3 Hl H o. O Φ cn H Φ & d Ω Φ P- d tr H d Ω Φ 03 d H

PJ P dd = Φ P- fV da H 13 fu d Φ vQ ^ d H a Φ PJ P- tr Φ Ω Φ PJ ι-i ≤ d tr d li P- Φ φ Φ li d ii Ω iQ ds: d fr P- Öd P- P- Φ Φ i- tad φ ιQ φ

Hi H ιQ 3 H ιQ d Ω P_ d J / rt d rt Φ d Φ d H H N φ d Ω Φ l-i z Φ _l. Φ X Φ d l = r Φ J Ω P Φ fr - P- N d ιC! d P- φ rt φ ι-i rt tr P- td ∑: P-

PJ d. -. Φ P ⁾ dd H er a ^j d P> Ω cn φ dd P- φ d Φ d cn N Φ φ o Φ d <α tr I- ¹ ds: P ^J § cn td d tr ddd α rt Φ n. d iQ φ P ddd

P- O Φ O: rr P- ιQ H- Cd P>: d rt H d rt d • d Φ N rt Φ rt P ⁾ H d S! dd cn PJ P- cn tr dd cn H φ d cn Φ fr ιQ ιQ cn Φ Φ d ^ d Φ d rt O: rt φ

• H Ω φ d cn O- d iQ rt i-i Φ Φ ω ω H d Φ Ω α O l-i N d d

1 Φ P- H Φ ^■ <d Φ dd P- Φ cn d Hl tr PJ 3 1 Φ ii

3 Φ cn P- 1 ri- 1 li 1 d P- • Φ φ cn P> 1-1 Ω 1 1 1 d 1 1 1 1 1 tr

The storage effort for a voice confirmation is about a power of ten higher than that of a reference speech sample. Last but not least, the presented method generally achieves a higher word recognition rate than with a single use (only one speaker) of the speech recognizer. The improvement in the word recognition rate is based on the expansion of the language reference base of a word by training with several speakers.

The inventive step lies in the use of a common vocabulary for all users of a speech recognition system, with one word being assigned the reference speech pattern by several users. The method requires the rejection strategy described above for voice training and for voice recognition.

The method is suitable for speech recognition applications with a restricted user group of more than one user. These are, in particular, applications with voice control and command input, but also with voice-controlled database access. Exemplary embodiments are voice-controlled telephones (voice-controlled dialing from a telephone book, voice-controlled control of individual functions, such as that of the answering machine) as well as other voice-controlled machines / devices such as household appliances, toys and motor vehicles.

Advantageous developments of the invention are specified in the subclaims.

An embodiment of the invention is explained with reference to FIGURES 1 to 8.

Claims

Patent claims

1. Method for the voice-controlled initiation of actions that can be carried out in a device by a limited group of users with the following features:

(a) Through the language associated with at least one user of the user group of the device, the device is trained for at least one operating mode selected by the respective user in at least one language training phase to be initiated by the user in such a way that

(al) at least one of these users enters at least one reference speech utterance that can be assigned to the action into the device for at least one action,

(a2) a reference language pattern is generated from the reference language utterance by language analysis, whereby in the case of several reference language utterances, the reference language pattern is generated if the reference language utterances are similar, (a3) the reference language pattern is assigned to the action, (a4) the reference language pattern is stored without reservation with the assigned action or is only saved if the reference speech pattern is dissimilar to the other reference speech patterns already stored that are assigned to other actions, (b) the respective user enters a recognition speech utterance into the device in a speech recognition phase for the operating mode of the device selected by the user a, (c) a recognition speech pattern is generated from the recognition speech utterance by speech analysis, (d) the recognition speech pattern is compared with at least part of the reference speech patterns stored for the selected operating mode in such a way that the similarity between the respective reference speech pattern and the recognition speech pattern is determined and a similarity ranking the stored reference speech patterns are formed based on the similarity values determined, ω N>N> P ¹ P ¹

Cπ o cπ O cn o Cπ o 3 α d ω φ S O N a 3 d M td 3 <! cn fr P- cn o Cn a n P- ιP a ω

• da cn a P- P- Φ d rt d 13 Φ ri ^¬ cn Φ PJ rt dd ri ^¬ Φ

Φ Φ ^» as: d φ Φ ' H oo 13 Φ Hi ^" dd H Φ Φ li tr P^ ¹ d ι-1 P- — rt 1 Φ ^ _* HH tt) rt cn Φ d t- tt)

HH rt- < φ rt H P- Ω H rt φ d tu rt P- P ¹ a tr J tr JH ιp Ω dd φ d tr Φ Φ d cn P. o cn P- O ^• — ' Φ φ a Ω tr ^ cn H li d φ a* ι-J dd 1 li d P- a ⁴ dd H CΛ cn H tu

Φ ^P c Hi CΛ a P- 13 h_j Φ ςx Hi d α 3 PJ ri ^¬ a Φ P ? Φ dd M cn tr Φ aad H P- Φ daa φ fr rt 13 a ri ^¬ cn

Hrtn,

_~> PJ 13 Φ d H p7 P ⁾ d φ dda N tt) t_ϊ Φ d P- Φ H Φ Φ H Φ tt) ⁾ &j t H d Φ PJ J CΛ &_ tr «-. cn ω CQ ι Φ dd tr da cn P- d N cn d tu ddad HJ rt d Ω 13 H cn Φ rt Hi 3 ιp d Φ rt Ω d ri ^¬ Ω Pt 3 d P- φ ; a Φ Ω P- ö" ^» < H φ cn Q d Φ o Φ CΛ t 3 Φ tr t tt) d Φ d' td d rt dd cn dp' cn < g cn J d 13 Φ d ii 0 O ri ^¬ P- φ H fr Φ a H d 3 φ H- n PJ 3 5 Ω O d Φ Ω HH rt ι φ li tt) Ω Hl P- Φ Hi Φ ιP cn dd rt ao 3 J dd tr H cn a- ¹ d PJ pj: NN Φ H a P ^J tr φ d Φ P- φ 3 CΛ cn er cn d P- Φ d P- £ J ω U3 rt N p PJ Ω rt Φ dp : d fr li P- rt li rt cn Φ rt rt P-H

PJ Ω Ω ri ^¬ P- Φ Φ dd Ω tr Φ H ι PJ rt Φ φ Φ φ ad cn Φ P- Φ 13 tr Φ N Φ

H ^> rt N 1-1 JN ¹ P> d H Φ h

P P- H cn tr Φ Φ E P- d= ii a P- cn r+ cn cn P ⁾ d cn a Φ Ω a N d fr

3 a Φ Φ cn H & φ d H Φ H P- Φ ι-i 13 [fci: d ιp 13 Φ Φ d s: Φ

-? φ cn <5 X d P- et cn li d H ad tr tu ι-i £? a Hi ι-i ii cn $ d P- ad φ a 13 O o Φ d P- X 13 d P- Φ a φ d P ⁾ d Φ OJ ri ^¬ da 1-1 Φ d rr Φ ö φ ii d φ φ rt Φ a • d ιP Ω P- ii Ω i: tt) cn o Φ a -fd

N PJ H d Φ d 5, H d uq ι-( d H P- a Hi _τ P- Φ iP P ¹ r tr rt Φ 3 d

Φ li Ω P_ P- s Φ a Ω P- cn Φ Φ O 3 Ω Φ pd φ Φ Hi ⁾ P- ιP

13 S tr Φ φ < d Φ tr PJ n <! cn rf- P ¹ d tr dd H Φ P- d cn

Φ P- Φ 3 a li oad 13 o rt CD ι cn > - PJ cn P- a H dn cn ι-i P rt MJ rt li Φ P ¹ cn H dd td Φ rt φ er d ri ^¬ Ω Φ N Φ a 13

3 ^P tr "» PJ cn Q cn PJ " _* PJ r Φ Φ P- P- tt) tr ddda N Φ

P- odd Φ N Ω α d CΛ P- Hi P ⁾ ii rt O d ii TN Φ d 3 tu co rt aaad ι Jjd cn P. d tr Φ d rt O Φ d (Λ d 1 φ to a cn HP ¹ Ω er a Φ J Φ tr rt Φ tr J Φ 3 H Φ d ι-( N ii rt d P- Φ φ 13 tu: 0 tr

Φ Φ li cn Φ tr cn PJ: PJ= lf Φ ad JN Φ d rt Hi 3 -i cn φ tu: r- ^> H 3 a Φ et ι-i Φ rt tr cn fr td P. P" d Φ ι dd H tu: cn Φ PJ J φ a NH Φ > v tr if H P- Ω g cn li d er P- dd J nd PJ d c_ cn Φ φ d P- tu: &5

M rt d Ω Φ d P- ¹ P- ddd P- cn O Hi Φ CΛ d P ⁾ Φ da* H ι rt φ tr P- 1-1 CΛ α tr da P- 1-1 ιp ddd cn Φ 13 φ H oo rt P- dd P P- lf

N a a Ω 13 cn l-i cn Ω Ω d rt rt ι-1 l-i a P" l-i Φ P- ιP N a d Ω φ J d

X a Φ tr H 3 Φ 5. 3 " tr Φ d N PJ Φ a PJ cn d ιP a P" Ω Hi cn Φ cn tr i-i d d

Φ H PJ PJ P- Φ PJ ι-i u3 Φ dt Φ Ω ri ^¬ Φ Φ d P ¹ tr O 13 H et fr fr cn ι d Ω to et H &o a ^ cn H cn Φ 3 P ⁴ tt) rt Φ φ P ¹ H Φ Φ PJ Hi tr 1 α ι a* Φ r Φ QP ⁾ 13 ιp d P d PJ rt P- ι JH P- d = tr

Φ g • 5 a rt N Hl d tr ιP φ a P- d P- d cn cn Φ Ω ϊ? rt d tr Φ i-i PJ d φ d P- d s: P- Φ d PJ φ Hi Φ d cn tr cn P- rt rt _t d PJ: cn rt H N

Φ ii J fr cn φ dd P- d fr rt _cn X d: rt P- rt φ cn Φ * P> § h- ¹ tr ii ^ tr Φ cn 1-1 rt ϊ fdd P- d PJ tu φ tr o P- Φ Φ td P- Φ Ω Φ Φ X tr H to Φ ι-{ d Cn Φ X cn Ω H" ds: H Ω d <! cn d ι-id H tr 1-1 ao φ rt φ H d φ d ri ^¬ dod rt tr P- ι φ Φ f^ φ 1 Ω rt d 3 cn φ rt a ii P- ri ^¬ Hi P- a tt ) P ⁾ a tr 1 Φ fr Ω Hi dd er r Φ J ₃ __;

•t S tr PJ Hi p- Ω d α d tt) ad H Φ Φ rt ι-1 Φ tr O d Φ o Φ N d Φ rt tr IS! Φ ^ιQ d X φ d Φ φ Φ d ι- ι-i tu: rt P- ι a cn

Φ Φ P- da Φ Φ Φ ιQ odd M tr tr ri ^¬ a td a ^• r cn cn Φ J er 13

P ^J •τ) P- tr d P- P-

P ¹ H Ω rt d Φ rt φ d Ω cn tr rt 1 Ω P- P- * lf Hi 3 rt P ⁾ JOP ⁾

P- o lP cn Φ p- rt Φ H a tr φ Φ P* Φ Φ P- d O d td d Ω φ -. tr Φ rt Φ tr • d P- P- 1 P- &3 a Φ > ι-i H- P ^J Ω ιP a H tr li i- I a 3 cn P- | Φ Φ Φ Ω P" P- v> Φ d tr Hi Φ Φ i^ ¹ ιP d P ⁾ φ d tr tr P- Φ er d φ 0 HH φ Φ d 3 cn cn Φ Ω P- N 1 d 1 ιQ 1 1 P tr 1 1 1 1

4. Method according to one of claims 1 to 3, characterized in that at least one word is entered by the user as a reference speech utterance.

5. The method according to one of claims 1 to 4, characterized in that the user assigns at least one user-specific identifier to the language training phases he has carried out.

6. The method according to one of claims 1 to 5, characterized in that the user input of several reference speech utterances belonging to a language training phase is automatically controlled by the device in that the end of the first entered reference speech utterance is recognized by the device by means of speech activity detection, because No further speech activity by the user that can be attributed to this reference speech utterance has taken place within a predetermined time, and the user is informed by the device of the time-limited input option for at least one further reference speech utterance.

7. The method according to one of claims 1 to 5, characterized in that the user input of several reference speech utterances belonging to a language training phase is controlled by interaction between the user and the device, in that the user informs the device through a specific operating procedure that several Reference language utterances are entered by him.

8. The method according to one of claims 1 to 7, characterized in that in different language training phases the users use different reference language utterances for an action, for example in different languages “German and English”.

9. Method according to one of claims 1 to 8, characterized in that the user enters information, e.g. a telephone number, through which the action is defined.

10. The method according to claim 9, characterized in that the information is entered using biometric input techniques.

11. The method according to any one of claims 1 to 10, characterized in that the information is entered before or after the reference speech utterance is entered.

12. The method according to any one of claims 1 to 11, characterized in that the action is specified by the device.

13. The method according to any one of claims 1 to 12, characterized in that the recognition speech utterance can be entered at any time in the speech recognition phase with the exception of during the speech training phase.

14. The method according to one of claims 1 to 13, characterized in that the recognition speech utterance can only be entered if the speech recognition phase in the device has previously been initiated by the user.

15. The method according to one of claims 1 to 14, characterized in that The language training mode is ended by saving the reference language sample.

16. The method according to one of claims 1 to 15, characterized in that the user is informed of the entry of an inadmissible recognition speech pattern.

17. The method according to any one of claims 1 to 16, characterized in that the speech recognition phase is initiated like the speech training phase.

18. The method according to any one of claims 1 to 17, characterized in that the voice-controlled initiation of actions that can be carried out in a device is carried out in telecommunications terminals.

19. The method according to one of claims 1 to 17, characterized in that the voice-controlled initiation of actions that can be carried out in a device in household appliances, in vehicles, in consumer electronics devices, in electronic devices for control and command input, e.g. a personal computer or a Personal Digital Assistant.

20. The method according to claim 17, characterized in that in a first operating mode of the telecommunications terminal the voice dialing from a telephone book or the voice-controlled sending of "Short Message Service" messages from a "Short Message Service" memory is carried out.

21. The method according to claim 17 or 20, characterized in that In a second operating mode of the telecommunications terminal, the voice control of functional units, for example answering machines, "Short Message Service" memory, is carried out.