Data processing: speech signal processing – linguistics – language – Speech signal processing – Application
Reexamination Certificate
1998-06-22
2002-05-21
Tsang, Fan (Department: 3645)
Data processing: speech signal processing, linguistics, language
Speech signal processing
Application
C704S270000, C379S088010
Reexamination Certificate
active
06393403
ABSTRACT:
FIELD OF THE INVENTION
The present invention relates to mobile communication devices and in particular to mobile communication devices arranged to recognise and act upon commands spoken by a user. The invention is applicable in particular, though not necessarily, to mobile telephones.
BACKGROUND OF THE INVENTION
It has long been desired to incorporate speech recognition functionality into mobile communication devices such as mobile telephones. By enabling a user to operate a telephone with spoken commands, e.g. to dial a particular telephone number, operations can be performed more quickly and more easily than is possible with conventional keypad arrangements. Furthermore, it becomes possible to dispense with all or a large portion of the keypad resulting in a considerable space say and allowing further miniaturisation of devices to take place. A similar desire exists to corporate speech synthesis functionality into mobile communication devices allowing requests and responses to be ‘spoken’ by the device to the user, thus simplifying the operation of the device.
The generic name for a system which comprises both speech recognition and speech synthesis functionality is ‘a speech dialogue system’ (SDS). In a SDS, speech recognition is usually accomplished using a set of ‘templates’, each template corresponding to one spoken command. A template is generated by analysing the corresponding command spoken by anything from one to several hundred different speakers. Typically, this analysis comprises extracting parameters, such a frequency spectrum coefficients, from each spoken version of the command and which are characteristic of the spoken command. Where a number of different speakers are used, the parameters are combined in an appropriate way to form a single set of parameters or template.
The generated templates are stored in a look-up table. When a command is spoken and converted into electronic form, the command is analysed to extract therefrom those same parameters which were used to form the templates. The look-up table is then searched to identify whether or not the table contains a template which contains the same, or similar parameters to those of the spoken command. The search may be conducted for example using a least squares measure whilst the degree of similarity required to flag a matching entry is predefined. If a matching entry is identified, then the action corresponding to the command is carried out by the device controller. If no matching entry is found, the user is notified to this effect and is asked to repeat the command.
Templates generated using only a single speaker are known as ‘speaker dependent templates’ as they can generally only be used to reliably recognise commands spoken by the person from whom they were generated. In the case of a SDS incorporated into a mobile telephone, this person is the telephone subscriber, whilst software (and hardware if necessary) for generating the templates is incorporated into the telephone.
‘Speaker independent templates’ is the term use to describe templates generated using commands spoken by a large number of different speakers. As speaker independent templates represent a kind of average of an entire class of speaker, they can be used with considerable reliability to recognise spoken commands independently of the speaker. In the case of a mobile telephone, speaker independent templates can be stored in a memory of the device prior to supplying the device to the end-user.
There are advantages and disadvantages to using either speaker dependent or speaker independent templates in mobile telephones. When speaker independent templates are used, the end-user does not need to perform a training or template generation operation, a time consuming and often tedious operation which must be performed when speaker dependent templates are used. However, in the case of speaker independent templates, it is still necessary to generate templates for storage in a memory of the phone prior to supplying the phone. The recording and analysing of commands spoken by several thousand speakers is an extremely demanding and time consuming exercise and can significantly delay the ‘time to market’ of a new product. This is especially so for a phone offered to a number of different language groups here a set of templates must be generated for each language which the phone supports. Furthermore, the provision in a telephone of even a few such sets of templates can present a considerable memory problem (tens or hundreds of Kbytes per language), increasing the size and/or cost of the telephone. This problem is compounded when speech synthesis functionality is included, it being necessary to store in the phone's memory a set of ‘responses’ for each supported language.
SUMMARY OF THE INVENTION
It is an object of the present invention to overcome or at least mitigate the above mentioned disadvantages of known speech dialogue systems for use in mobile communication devices.
According to a first aspect of the present invention there is provided a mobile communication device having speech recognition functionality and comprising:
a memory for storing a set of speech recognition templates corresponding to a set of respective spoken commands each template corresponding to a single human language;
transducer means for converting a spoke command into an electrical signal;
signal processing means arranged to analyse a converted spoken command together with templates stored in said memory to identify whether or not the converted spoken command corresponds with one of said set of spoken commands; and
means for receiving via a wireless transmission channel said set of templates and for storing this in said memory.
Preferably, said set of templates is a set of speaker independent templates. More preferably the set corresponds to a single human language. The memory may be arranged to store a plurality of sets of templates corresponding to respective human languages and received via said wireless transmission channel. The device may comprise selection means for enabling the user to select a language set for reception via said wireless transmission channel and/or to select a language set for use by said signal processing means.
The present invention makes it possible to provide a mobile communication device having speech recognition functionality, for example a mobile telephone, to an end-user with no (or only a limited number of) speaker independent language template sets. The end-user may then ‘download’ one or more selected sets of speaker independent templates from a central station. This overcomes the memory storage problem encountered when it is necessary to supply the telephone to the end-user with a set of templates for each supported language, prestored in a memory of the phone. The present invention allows a large number of languages to be supported in a speaker independent manner as there is practically no limit to the number of sets of templates which can be stored at the central station. Furthermore, as sets of templates become available for additional languages, these can be made available to the end-user via the telephone network.
A device according to the above first aspect of the invention may comprises means for generating a set of speaker dependent template using commands spoken by a user and converted by said transducer means, and mean for storing templates so generated in said memory of the device. The device may further comprise means for selecting either the set of speaker dependent templates or a set of speaker independent templates, where such a set is stored in the memory, for use by said signal processing means.
Devices embodying the present invention may comprise means for receiving via said wireless transmission channel a set of speech responses and means for storing such a received set in said memory. The set of received responses may correspond to a single human language.
According to a second aspect of the present invention there is provided a mobile telephone network comprising a multiplicity of mobile communication devices,
Nokia Mobile Phones Limited
Opsasnick Michael N.
Perman & Green LLP
Tsang Fan
LandOfFree
Mobile communication devices having speech recognition... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Mobile communication devices having speech recognition..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Mobile communication devices having speech recognition... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2863650