DE19742369A1 - Digital audio signal recording and reproduction device - Google Patents

Abstract

The device includes a processor (DSP) for control of the recording and reproduction processes, an analogue interface (ANL) for recording the analogue audio signals (AR), and a coupled analogue to digital converter unit (AID) for digitising the analogue audio signals. There is a digital interface (DIG) for recording the digitized audio signals (AD). The device comprises at least one connector for the interchangeable writable semiconductor memory cards (S1... S3) for recording digital audio information (AD(t)), and a D/A converter unit (D/A) for reproduction of the audio information stored on the semiconductor memory cards. A microphone (M) is preferably provided for entering speech information which is digitised by the A/D converter unit.

Description

The invention relates to a device for recording and how output of audio signals.

Audio signals, which be essentially music recordings traditionally meet Ma gnet tapes, so-called audio cassettes, copied. A thief dubbing compact discs is a common process (CD's) or radio broadcast on audio cassettes. The audio signals are usually recorded analogously.

However, digital recording and playback are also known devices with magnetic tapes that are on the market, for example, under DAT (Digital Audio Tape) are available.

However, such magnetic cassette tapes have considerable che mechanical disadvantages. The quality of an analog on drawing is bad and access to a certain one Spot is slow and extremely uncomfortable. In addition, such devices wear out quickly and interference problems with the tape winding occur frequently.

Other known digital audio recording and playback devices store audio signals on diskettes or other magnetoopti cal disk. An example of this technology is technology known as "Mini-Disk". This will make access faster and more convenient to certain areas of audio information allows however, occur mainly due to mechanical loads such as vibrations, often problems such as wow and flutter conditions or dropouts during playback due to Interruptions in data transfer from the floppy disk.  

The object of the present invention is therefore a digita to create the audio recording and playback device which with regard to the operation and storage of audio signals len low wear and a low error and Failure rate and also a simple and enables convenient operation.

According to the present invention, this object is achieved by a device for recording and reproducing audio signals with the following elements:

• - A processor to control the recording and playback operations,
• - a digital interface for recording already digita lized audio signals,
• - an analog interface for recording analog Au diosignals with a downstream analog-digital Converter unit,
• - At least one slot for interchangeable writable re semiconductor memory cards for digital Au dio information,
• - A digital-to-analog converter unit for playing on Semiconductor memory cards stored audio information.

By putting the digital audio information on in this way exchangeable semiconductor memory cards is written and by these cards can be read again is a simple one Copy audio signals (e.g. from music CDs) as well as a Playback in CD quality possible. That way they can disadvantages of audio magnetic tape cassettes and disks can be overcome. The digital Au dio recording and playback device according to the present Er invention is preferably a battery or rechargeable battery operated Device with which a recording is either analog from an analog output device (e.g. a stereo system) or can take place via the intended digital interface,  if such is also offered by the recording source. Conventional personal computers, which with a sound card and are equipped with a CD-ROM drive, for example instruct a digital interface from which an audio signal can be digitally removed from audio CDs. A high how the delivery quality can be achieved by a device according to the In particular, the invention can be guaranteed because the audio information is stored digitally, where taking into account the available memory depending on the requirement, the resolution is very high (e.g. 16 Bit or 24 bit / technology) can be selected. Also at the recording of analog audio signals and their digitali sation by means of the analog-digital converter unit can last tere depending on the application with different sampling rates (e.g. 22 500 Hz or 44 100 Hz) can be selected. same for for the resolution of the audio signal (16 bit, 24 bit etc.).

So that the sampling also all essential audio signal recorded, the sampling theorem must be fulfilled. This states that the highest frequency occurring in the signal is small must be less than half the sampling frequency, otherwise it will due to the so-called aliasing effect the signal information can come.

A first advantageous embodiment of the device according to The present invention has a microphone element for surrender of speech information, the Sprachinformatio can be digitized using the analog-digital converter unit are.

This opens up the possibility that the user at the recording of audio signals the title of the recorded Piece (song, text, etc.). This about that Voice information provided by the microphone element is via the already mentioned analog-digital converter unit digitali  based and also on the semiconductor memory card wrote. When searching for a title or when selecting it a title of audio information, which is on the half lead ter card, this spoken text, etc. then be output as an audio signal.

In a further advantageous embodiment of the device according to the present invention is also a memory cherverwaltungselement provided with which the digitized Voice information already on the semiconductor memory card stored digital audio information can be assigned can and assigned in this way also on the half conductor memory card can be saved.

The memory management can in particular be constructed that every recorded piece of music etc. has the associated recorded text is assigned. Access is then made optionally at the beginning of a certain piece of music or even within one piece. Since a semiconductor memory ver is used, the access for the user is almost complete without hesitation.

The recorded pieces of music become blue, for example fend filed. A piece is selected first via a number (comparable to an audio CD).

In a further advantageous embodiment of the device according to the present invention is a speech recognition engine dul provided with that contained in the voice information Tax commands are recognizable.

The fact that the digital recording and playback device Includes speech recognition module, for example with the voice information provided by the microphone element can be evaluated is a simple and effective loading  serving the digital audio recording and playback device tion enables. In this way, broken tax mandos of the user can be recognized. A control command, e.g. B. the keyword "title", can list the recorded arrange a piece of music. In such a case the device the order of the memory numbers with the addition listen to the recorded information of a speech signal spent. Other control commands can, for example, the Number of a track to be played, a control command "Next" for the selection of the next title, a tax com mando "stop" to stop playback, a control command "Pause" etc.

In a further advantageous embodiment of the device according to the present invention uses speech recognition a new module for classifying voice information ronal network.

Speech recognition is a pattern recognition task, which is divided into two stages. Which he The first level concerns feature extraction, the second level Classification. The feature extraction is used for the Recognize relevant features from the language information on ex trah and eliminate relevant parts. In the class Sification is then based on the current extracted characteristics concluded on the meaning of the speech information.

A particularly suitable and effective method for remembering Painting extraction and classification is in the international Patent application WO 93/02448 called "Method and arrangement for recognizing single words of spoken words Language "has been suggested. Regarding details of a possible Chen implementation of the speech recognition module is therefore based on referred to the font mentioned.  

A special advantage resulting from the use of a neuro nal network for the classification of voice information, is that this is application specific to this application basic suitable control commands can be trained.

In a further advantageous embodiment of the device According to the present invention, the digital audio are formations and voice information in a compressed form the semiconductor memory card can be stored, which has the consequence that comparatively more audio information for the same Storage space can be stored on the semiconductor memory cards are. All are used to compress digital audio signals Usable compression methods, particularly advantageous is responsible for the device according to the present invention the Atrac III algorithm used in mini-disk technology used.

A further advantageous embodiment of the device according to The present invention is characterized in that the Memory management element and the speech recognition module are implemented using the processor. In this way the performance of the processor can be better use and thus the effort and thereby the costs in the realization device according to the present Er further reduce the invention.

In a further advantageous embodiment of the device according to the present invention is a circuit arrangement of the components presented above, which is a particularly effective realization of the invention modern device allows. This is achieved that the processor is designed as a digital signal processor, which via a bus system consisting of data, address and Control lines, with a program memory, a working memory which at least one semiconductor memory card around  as well as with the digital interface, the analog Digital converter unit and the digital-to-analog converter unit connected is.

In a further advantageous embodiment of the device according to the present invention come several times bare semiconductor memory cards are used. This owns the The advantage that In. Stored on the semiconductor memory cards formations can be corrected at any time. In addition, parts of the audio information are exchanged for others without the entire content or the whole semblance memory card itself must be replaced. Especially suitable semiconductor memory cards are those in one EEPROM or Flash ROM technology.

According to a further advantageous embodiment of the device According to the present invention, training takes place the neural network using an external computer. There through the neural network on the detection of applika tion-specific control commands can be optimized. The at a knowledge base for the neuro nale network (training information) is then shared with the audio information also provided for specific applications tion on the digital interface of the invention Device provided and from this on a corre sponding corresponding semiconductor memory card. It is special advantageous, the training information for the interchangeable neural network on semiconductor memory cards increase.

A device according to the present invention in a portable music player deploy. Such devices have been used in the past The name Walk-Man became known, which, however, is the one verses described above with the corresponding disadvantages  Insert the audio magnetic tape cassette. Next to it are por table CD player known, but has the disadvantage zen that no recording of digital audio information is possible is. Both known portable music players is common that they are complex and expensive mechanical loading need controls. Such mechanical controls are subject to great wear and tear and are very faulty due. These also have the disadvantage that with portable Music players often encounter vibrations the mechanical elements do not always function correctly guarantee. By operating a portable music game device with a device according to the present Er Finding by voice commands can be time-consuming and expensive there are no mechanical controls. Since the music tracks from the Be servants whose names can be selected no longer apply additional effort for corresponding mechanical controls elements, which further reduces the susceptibility to errors can. Also a time-consuming labeling of the memory cards is unnecessary, which is also advantageous, because memory cards are now half a te have phone card and therefore only a label is difficult to do.

In this way, the handling of such a device and the semiconductor memory card both when recording and extremely comfortable when playing audio information bel. At the same time, the recording and playback quality according to the wishes and requirements by suitable Choice of sampling rate and resolution in digitization ensures. Since no mechanical elements are necessary, there is no wear and the error and failure rate is accordingly low.

Under the above conditions, a pre is recommended direction according to the present invention also for use  as an interactive toy, for example arranged in an egg a soft toy. In such a case, the semi-lead can useful audio information must be stored in memory cards, which are played on a corresponding control command become. This is an interactive communication a child with his toy.

Further advantages and inventive details emerge based on the following advantageous embodiment in Connection with the figures. Show it:

Fig. 1 is a schematic diagram of the circuitry of the digita len audio recording and reproducing apparatus according to the present invention,

Fig. 2 is a schematic diagram of the speech recognition module within the digital audio recording and playback device for recognizing and evaluating speech information.

The schematic diagram according to Fig. 1 shows a possible structure of the apparatus for recording and reproducing audio signals according to the present invention. For this purpose, a microprocessor, preferably a digital signal processor DSP, via a bus system B consisting of data, address and control lines, with a program memory element ROM, a RAM, which comprises at least one semiconductor memory card S1 to S3, and one digital interface DIG, an analog-to-digital converter A / D and a digital-to-analog converter D / A. Via the analog-digital converter unit A / D, analog audio information A _A (t), which is acquired via an analog interface ANL, is digitized.

Furthermore, a microphone unit M is provided, for. B. a Crystal or condenser microphone, its analog output  signal also supplied to the analog-to-digital converter unit A / D leads and can be digitized by this.

Furthermore, audio information A _D (t) which is already digitally available can be supplied to the device via the digital interface DIG mentioned. Digital audio information A _D (t) which has been converted in this way from analog to digital or is acquired directly via the digital interface DIG is advantageously stored in a compressed manner on semiconductor memory cards S1 to S3. Such semiconductor memory cards are connected, for example, via corresponding plug contacts to the Bussy stem B and can be replaced in this way. Advantageously, rewritable semiconductor memory cards are used, especially in EEPROM or FLASH technology.

The program memory ROM contains a suitable program for operating the microprocessor DSP and the other components already mentioned. The further provided digital-to-analog converter unit, which is also connected to the bus system B, is used to convert the digitally available audio information A _D (t) into audible analog signals, which can be reproduced via a loudspeaker or headphones, etc.

In the illustration according to FIG. 2, one possibility of realizing the speech recognition module is shown. This speech recognition module can be realized particularly advantageously with the aid of the digital signal processor DSP and a suitable programming in the program memory ROM. For this purpose, for example, audio information A _D (t), which is acquired and converted to analog / digitally converted and is stored in the form of digital amplitudes present over time t, is transformed into an amplitude spectrum SP in the frequency range f. Fourier transform methods such as the Discrete Fourier Transform DFT or a Fast Fourier Transform FFT are particularly suitable for this (cf. in particular Schrüfer E .: Signal processing - numerical processing of digital signals, Munich, Vienna, Hanser-Verlag 1990, page 147 ff.).

With the help of the amplitude spectrum SP thus determined extracted suitable features E1 to E3, which a neuro nn network can be fed. The necessary scope of neural network NN is characterized by the complexity of the evaluating language information. The number of inputs elements (dimension of the input feature vector) determines the Number of network inputs. The number of network outputs A1, A2 is determined by the number of patterns to be recognized. The necessary number of input neurons N1 to N3 is from the number of patterns or individual words to be recognized and the dimension of the input feature vector and sub distinguished between the patterns to be recognized or individual words dependent. The number of input neurons N1 to N3 leaves best by means of an iterative configuration process from several so-called learning and training phases of the neural network. The ones to be recognized Patterns or single words to the inputs of the neural network works and with the desired output or Control command compared. The comparison is based on the so-called back propagation algorithm statements about the Classification quality of the individual neurons N1 to N5 obtained (see also Werbos, P., Beyond Regression: New Tools for Prediction and Analysis in the Behavioral Sciences. Ph. D. thesis, Harvard university, Cambridge, MA, Aug. 1974).

Each neuron N1 to N5 consists of a summing element, the the values at the inputs of the neuron are weighted be supplied, as well as a non-linear activation radio tion, e.g. B. the so-called sigmoid function, between the  Summer and the output of the neuron. Training one Neurons N1 to N5 is done by determining the weights, with which the input values are amplified. Aim of the trai ning phase of the neural network NN is thus the Festle weighting of a network in such a way that afterwards the desired recognition process takes place. The weights are the carriers of the information of a neural network works and represent their knowledge base. If the goal ei correct recognition process during a training phase If this is not achieved, the network can be adapted topology may be required. As a rule, the number of Neurons are increased. Then the training process with an enlarged neural network like this repeats.

In the classification phase, the input side then opens the manner described above features E1 to E3 and the neural network NN, wel The correspondingly classified tax on the output side commands SK for further processing. On the se way can be detected in the microphone element M. th voice information control commands SK recognize which again to operate the device according to the invention are suitable.

Possible applications of this device according to the invention are, for example, in the storage of information of speech works in the form of digital audio information Semiconductor memory cards. Textbooks are examples of this. The device according to the invention reacts accordingly the control commands SK of the user, matching the Interactive and voice-controlled information in the textbook Text, music, questions and associated answers are output can be.  

Another advantageous application of the fiction The device consists of a portable music player advises with which classic radio play cassettes are replaced and with additional properties can be added (radio play with quizzes etc.). It is particularly advantageous that a sol portable music player with no mechanical loading controls and thus significantly fewer errors lig than conventional music players. Different Music titles can be, for example, at the beginning recorded way on semiconductor memory cards men or through already described semiconductors replace memory cards using the plug contacts mentioned.

Application-specific can also be made using a conventional Chen personal computer or via a data network, such as the Internet, information about the digital interface DIG on the rewritable semiconductor memory cards be loaded and further application specific information such as training information mation or knowledge bases for the neural network NN.

Furthermore, it is conceivable that the device according to the lying invention as a separate speech recognition module in Use connection with a personal computer, which connected to each other via the digital interface DIG are. This has the advantage that the performance the personal computer is not charged and this for others applications are still available.

The use of the invention is also conceivable Device in an interactive toy. For this, the Device, for example, inside a soft toy be classified. The device responds to control commands of the operator / child with the reproduction of this tax command matching, stored on semiconductor memory cards  Reply. The repertoire of answers with the corresponding ones the control commands can be via different semiconductors memory cards are exchanged.

In addition, the device according to the present invention in various other technical devices, such as telephones, audio devices, televisions, video devices but also household appliances such as Washing machines etc. are installed and thus with egg A cheap and effective way to recognize speech or expand voice output.

Claims (12)

1. Device for recording and reproducing audio signals with the following elements:

• - a processor (DSP) for controlling the recording and playback processes,
• - a digital interface (DIG) for recording already digitized audio signals (A _D (t)),
• - an analog interface (ANL) for recording analog audio signals (A _A (t)) with a downstream analog-digital converter unit (A / D),
• - at least one slot for interchangeable writable semiconductor memory cards (S1... S3) for recording digital audio information (A _D (t)),
• - A digital-to-analog converter unit (D / A) for reproducing audio information stored on semiconductor memory cards (S1... S3).

2. Device according to claim 1, characterized ge indicates that a microphone element (M) for on Giving speech information is provided, which by means of Analog-digital converter unit (A / D) can be digitized. 3. Apparatus according to claim 2, characterized in that a memory management element is provided with the digitized speech information on semiconductor memory cards (S1 ... S3) digital audio information (A _D (t)) can be assigned and stored. 4. Apparatus according to claim 2 or 3, characterized characterized in that a speech recognition module is provided with that contained in the speech information Control commands (SK) are recognizable.   5. The device according to claim 4, characterized ge indicates that the speech recognition module for Classification of voice information a neural network (NN) uses which trai on suitable control commands (SK) nable. 6. Device according to one of the preceding claims, characterized in that digital audio information (A _D (t)) and voice information in compressed form can be stored on the semiconductor memory cards. 7. Device according to one of the preceding claims, characterized in that the memory administrative element and the speech recognition module by the Processor (DSP) are realized. 8. Device according to one of the preceding claims, characterized in that the processor is pronounced as a digital signal processor (DSP), which via a bus system (B) consisting of data, address and control lines, with a program memory (ROM), a working memory (RAM), which has at least one semiconductor memory card (S1... S3) includes, as well as with the digital interface (DIG), the analog-to-digital converter unit (A / D) and the Digi tal-analog converter unit (D / A) is connected. 9. Device according to one of the preceding claims, characterized in that each half lead memory card (S1 ... S3) can be written to multiple times. 10. Device according to one of claims 5 to 9, there characterized by that over the digital Interface (DIG) audio information and training information for the neural network (NN) application-specific can be provided by an external computer.   11. Portable music player with one device after any of the preceding claims. 12. Interactive toy with a device after one of claims 2 to 10.