EP4403830A1 - Cooking appliance and method - Google Patents

Abstract

Cooking appliance (10), with the following features: a heat source (12) for heating a medium or a cooking vessel; means for sound detection (54), which are designed to receive an acoustic signal originating from the medium or the cooking vessel and/or to convert it into a digital acoustic signal; a processor (55), which is designed to analyze the digital acoustic signal with regard to one or more features and to determine a cooking state for the medium based on the one or more features.

Description

Embodiments of the present invention relate to a cooking appliance, cooking accessories and a method. Preferred embodiments relate to a machine learning-based air/structure-borne sound sensor with associated software solution for safe, energy-efficient and automated cooking, frying and warming up.

Temperature measurements are irrelevant to humans when cooking. Whether it's 71°C on Mount Everest or 100°C on the Atlantic: water boils when it bubbles; fat is hot when it sizzles. In this way, even an inexperienced cook can monitor several pots by "listening" and make adjustments on the stove. The human ear is able to both recognize individual sounds and put them into the right context.

This "listening" should be automated to monitor and improve the cooking process. Currently, water boiling over or burning fat is not detected reliably or in a user-friendly manner. The stove continues to heat and if no one is present to manually regulate the temperature/heating power/power level, this can potentially result in an accident. Furthermore, more energy is often added to the cooking process than necessary, e.g. due to a delayed power adjustment and the premature reaching of the maximum possible water temperature (boiling temperature).

The state of the art already includes some approaches for monitoring cooking, frying and/or boiling processes. The bibliography contains a collection of patent applications from this field.

• ∘ Beispielsweise werden Infrarot-Temperatursensoren in der Dunstabzugshaube integriert, welche jedoch nur ohne Topfdeckel funktionieren.
• ∘ Des Weiteren werden Temperatursensoren in die Herdplatte eingesetzt. Diese reagieren jedoch verzögert und liefern deshalb keinen Schutz gegen überkochende Flüssigkeiten.
• ∘ Teure, externe Temperatursensoren (z. B. Siemens Kochsensor Plus) müssen per Hand an einem speziellen Topf angebracht werden. Zudem müssen Batterien manuell aufgeladen und die Funkverbindung manuell hergestellt werden.

Currently available high-level stove systems rely on temperature sensors to monitor the cooking and frying process. Accordingly, existing solutions differ mainly in the position of the attached sensors; each position has disadvantages:

• ∘ For example, infrared temperature sensors are integrated into the extractor hood, but these only work without a pot lid.
• ∘ Temperature sensors are also installed in the hotplate. However, these react with a delay and therefore do not provide protection against liquids boiling over.
• ∘ Expensive, external temperature sensors (e.g. Siemens Cooking Sensor Plus) have to be attached to a special pot by hand. In addition, batteries have to be charged manually and the radio connection has to be established manually.

Humidity sensors for monitoring water vapor are also possible. Other systems (see scientific reference) focus on gas stoves to detect fires, e.g. using gas sensors to monitor gas leaks or IR sensors above the stove top for early fire detection. This usually involves feedback to the gas supply of the stove or directly to the user via an app.

However, as the above points show, relying solely on temperature readings only works when used correctly or when using special pots, tolerating time delays, or only for gas. This is not guaranteed for all users. Therefore, there is a need for an improved approach.

The object of the present invention is to provide a concept for monitoring a cooking process (including frying, warming up, etc.) which overcomes the disadvantages present in the prior art and in particular offers an improved compromise between cost efficiency, monitoring functionality and monitoring accuracy.

The problem is solved by the subject matter of the independent patent claim.

Embodiments of the present invention provide a cooking appliance (e.g. a hob) with a heat source (e.g. an induction plate) for heating a medium (e.g. food to be cooked, such as vegetables or meat or cooking water or frying fat) or a cooking utensil (e.g. the pot). The cooking appliance also comprises means for sound detection (e.g. airborne sound or structure-borne sound) which are designed to receive an acoustic signal (e.g. a boiling sound) originating from the medium or the cooking utensil and/or to convert it into a digital acoustic signal. According to embodiments, these means for sound detection can be, for example, an airborne sound or structure-borne sound microphone or generally a microphone. Furthermore, the cooking device comprises a processor which is designed to analyze the (received and digitized) digital acoustic signal with regard to one or more features and to determine a cooking state (e.g. a boiling state) for the medium based on the one or more features or raw data.

According to one embodiment, the cooking appliance can have the form of a hob. There is no restriction to certain types, such as a gas hob, induction hob or infrared hob. According to one embodiment, the cookware can be a pot or a pan. The medium can, for example, be a food to be cooked, e.g. meat, or a cooking medium, e.g. cooking water or fat. According to one embodiment, the cooking appliance has the form of a hob that comprises several hotplates or cooking positions, wherein the processor is designed to carry out cooking position detection or hotplate detection based on the digital acoustic signals and/or additional information. The hotplate position or the hotplate detection information represents one of the above-mentioned features.

Embodiments of the present invention are based on the knowledge that a cooking state can be monitored directly by acoustic monitoring by analyzing the recorded acoustic signal (structure-borne sound signal or airborne sound signal) with regard to one or more characteristics. This makes it possible to make much more detailed statements about the cooking or kitchen processes or even additional information about the presence of people in the kitchen/cooking area. Such a sensor system reacts much faster than existing sensors, so that changes in the cooking state can be responded to quickly. The monitoring is also less dependent on cookware and/or open or closed lids of the cookware.

According to embodiments, the cooking status is detected only on the basis of the acoustic signals, i.e. without additional information from the temperature signal or similar. The currently available signal can also be sufficient to detect the cooking status without taking the previous signal curve into account.

According to a further embodiment, a cooking accessory is created, e.g. as a supplement to a cooking device. The cooking accessory can, for example, comprise means for sound detection and the processor. The cooking accessory is intended for use for a cooking device with a heat source for heating a medium or a cooking vessel. and includes the two units already explained above. The advantage here is that the system can also be attached to existing kitchen appliances and these existing kitchen appliances can be expanded, so to speak.

• Temperaturinformation des Mediums;
• Siedeinformation eines Fluids als Medium;
• Verdampfungs- und/oder Verbrennungszustand eines Fettes als Medium;
• Garzustand des Mediums und/oder des Kochguts als Medium.

According to a further embodiment, a differentiation between one or more cooking states can be made based on one or more patterns of features, ie on the basis of a combination of features. In particular, predefined cooking states can be recognized and differentiated. According to one embodiment, the cooking state can come from the group of cooking states that include the following:

• Temperature information of the medium;
• Boiling information of a fluid as a medium;
• Evaporation and/or combustion state of a fat as a medium;
• Cooking status of the medium and/or the food being cooked as a medium.

The means for sound detection can, for example, be directly assigned to a cooking utensil or the medium, so that there is then also a direct assignment to the cooking position. According to embodiments, it would also be conceivable for the means for sound detection to be aligned with the one or more hotplates or to align themselves automatically. The exact arrangement depends on the implementation of the means for sound detection. According to embodiments, these can have means for sound detection of airborne sound, one or more microphones and/or means for sound detection of structure-borne sound. In the case of the means for airborne sound, the microphone is advantageously aligned accordingly with the cooking utensil and/or the means to be monitored. In the case of means for structure-borne sound monitoring, the means for structure-borne sound detection can be arranged directly on the cooking utensil. Furthermore, according to embodiments, means for sound detection can also comprise external means for sound detection that are arranged directly on the cooking appliance or the cooking utensil or the kitchen ceiling in the area of the hob or the extractor hood, etc.

• A/D-Wandlung der akustischen Signale;
• Vorverarbeitung, welche ggf. Schritte wie Quantisierung oder Normalisierung beinhaltet;
• Ggf. Merkmalsextraktion, dies kann in Form einer Zeit-Frequenz-Transformation oder anderer vordefinierter Merkmale erfolgen; alternativ könnte auch mit direkten Zeitdaten (Rohdaten) als Eingabe für das ML Modell gearbeitet werden;
• Ggf. Signalfilterung, insbesondere zur Störgeräuschminimierung oder Minimierung von Störgeräuschen, herrührend von der Heizquelle (Induktion), beispielsweise in Form eines Bandpassfilters; oder
• Beamforming-Algorithmus zur Ortung bzw. Positionsdetektion oder Kochzonendetektion.

According to one embodiment, the processor is configured to perform one or more of the following procedures:

• A/D conversion of the acoustic signals;
• Preprocessing, which may include steps such as quantization or normalization;
• If necessary, feature extraction, this can be done in the form of a time-frequency transformation or other predefined features; alternatively, direct time data (raw data) could be used as input for the ML model;
• If necessary, signal filtering, in particular to minimise noise or noise originating from the heat source (induction), for example in the form of a bandpass filter; or
• Beamforming algorithm for location or position detection or cooking zone detection.

This means that the processor can have an A/D converter or can be designed to carry out preprocessing, for example by means of steps such as quantization or normalization. According to embodiments, the processor can be designed for feature extraction, for example based on a time-frequency transformation, or another type of feature extraction. According to further embodiments, the processor can have a filter for signal filtering, in particular for minimizing noise. This filter advantageously enables the minimization of noise originating from the heating source (induction), etc. One possible implementation variant would be a bandpass filter.

According to a further embodiment, the processor is designed to carry out the analysis using machine learning, for example in the form of a neural network. For example, the processor can train the neural network and/or the ML algorithm using training data and/or using a current digital acoustic signal or using several digital acoustic signals (for example together with associated cooking status information). The processor is designed, for example, to be trained on site and/or to be trained in the delivery state and/or to be adapted on site and/or to be trained using federated training.

According to a further embodiment, a digital controller is created which determines the heating power for achieving a target cooking state depending on the current cooking state.

At this point, it should be noted that according to embodiments, the processor is designed to post-process a signal derived from the digital acoustic signal by means of privacy enhancement processing, e.g. by means of a speech filter. This advantageously enables acoustic monitoring without the privacy-relevant information being passed on.

• Anregungsfrequenz des Kochgeschirrs;
• Temperatur des Kochgeschirrs;
• Topfpositionsinformation;
• Topferkennungszustand;
• Kochgeräteinstellung;
• Information über das Medium (Gargut und/oder Garmedium).

According to further embodiments, the processor can also receive additional information in order to carry out the analysis. Thus, the processor can be designed to carry out the analysis based on further features, e.g. features from the group comprising the following:

• Excitation frequency of the cookware;
• Temperature of the cookware;
• Pot position information;
• Pot detection state;
• Cooking appliance setting;
• Information about the medium (food and/or cooking medium).

The list is not exhaustive. Combinations are also conceivable. In other words, this means that, according to the embodiments, the cooking appliance can advantageously be expanded to include additional sensors or can access additional sensors. As an example, it should be noted that the combination of acoustic monitoring with temperature measurement offers advantages, particularly for the learning process. Conversely, it is said that temperature measurements generally already provide useful information on the cooking status, but this can be valuablely supplemented by auditory noise monitoring.

According to one embodiment, the cooking appliance has an internal database and/or an interface to an external database, which the processor can access, for example, for analysis. The internal database or the external database comprises information relating to one or more reference features associated with a cooking state (in stored form), wherein the processor is designed to carry out an analysis using the one or more reference features.

According to a further embodiment, the cooking appliance can have a control system that is designed to regulate the heat source depending on the determined cooking state, in particular a power for the heat source. For example, the heat source can be reduced when a cooking state is reached, in particular a boiling state.

• Empfangen eines akustischen Signals herrührend von dem Medium oder dem Kochgeschirr mittels Mitteln zur Schalldetektion und/oder Überführen in ein digitales akustisches Signal;
• Analysieren des digitalen Signals im Hinblick auf die ein oder mehreren Merkmale und Ermitteln eines Kochzustands anhand der ein oder mehreren Merkmale.

According to a further embodiment, a method for monitoring a cooking appliance with a heat source for heating a medium or a cooking vessel can be created. The method comprises

• Receiving an acoustic signal originating from the medium or the cookware by means of sound detection means and/or converting it into a digital acoustic signal;
• Analyzing the digital signal with respect to the one or more characteristics and determining a cooking condition based on the one or more characteristics.

According to one embodiment, this method can also be computer-implemented. In this respect, a method refers to a computer program for carrying out this method.

Fig. 1A-1B

schematische Darstellungen eines Kochgeräts zur Erläuterung der Funktionalität bei Ausführungsbeispielen;

Fig. 1C

schematische Diagramme zur Illustration der Steuerung der Leistung bei Ausführungsbeispielen;

Fig. 1D bis 1G

schematische Darstellungen zur Illustration der Regelung der Temperatur gemäß Ausführungsbeispielen;

Fig. 2A

ein schematisches Blockdiagramm zur Illustration der Integration von Mitteln zur Schalldetektion in einer Regelstrecke für das Kochgerät gemäß Ausführungsbeispielen;

Fig. 2B

eine schematische Illustration zur Erläuterung unterschiedlicher Merkmale bei der Analyse gemäß Ausführungsbeispielen;

Fig. 2C

eine schematische Übersicht zur Erläuterung von möglichen Implementierungsvarianten für Mittel zur Schalldetektion gemäß Ausführungsbeispielen;

Fig. 2D

eine schematische Darstellungen (Fig. 2Dlinks, 2Drechts) zur Erläuterung möglicher Positionierungen der Mittel zur Schalldetektion bei Kochgeräten;

Fig. 2E und 2F

schematische Diagramme zur Erläuterung und Detektion von Störgeräuschen, die gemäß Ausführungsbeispielen filterbar sind;

Fig. 2G und 2H

schematische Illustrationen zur Erläuterung der Filterung von Störgeräuschen gemäß Ausführungsbeispielen;

Fig. 3A und 3B

schematische Darstellungen zur Erläuterung der Klassifizierung von Kochzuständen gemäß Ausführungsbeispielen;

Fig. 3C-3G

schematische Darstellungen zur Erläuterung von Anwendungen gemäß Ausführungsbeispielen;

Fig. 4

eine schematische Darstellung eines Kochgeräts mit Mitteln zur Schalldetektion sowie einem Prozessor gemäß einem Basisausführungsbeispiel des Kochgeräts;

Fig. 5a

illustriert schematisch die Ausführung des Kochgeräts bzw. Sensors als cyber-physisches System zur Illustration von Ausführungsbeispielen;

Fig. 5b

illustriert schematisch die Einordnung von Standard-Garmethoden nach Temperatur (bei Normaldruck) und Hörbarkeit zur Erläuterung von Ausführungsbeispielen;

Fig. 5c

illustriert schematisch charakteristische Merkmale von Kochgeräuschen, dargestellt als Mel-Spektrogramm zur Erläuterung von Ausführungsbeispielen;

Fig. 5d

illustriert schematisch einen exemplarischer Aufbau und Anwendung eines CNN am Beispiel von vier Kochzuständen gemäß Ausführungsbeispielen;

Fig. 5e

illustriert schematisch einen Versuchsaufbau und Hardware für die Datenerfassung im Rahmen der Machbarkeitsstudie zur Erläuterung von Ausführungsbeispielen;

Fig. 5f

illustriert schematisch ausgewählte Mel-Spektrogramme (200x200) für jeden Kochzustand aus dem Trainingsdatensatz (Betrachtungszeitraum innerhalb eines Spektrogramms beträgt 2,14 s auf der Ordinate, die Anzahl der Mel-Filter-Bänder beträgt 200, bei einer Bandbreite von 200 Hz bis 20.000 Hz) zur Erläuterung von Ausführungsbeispielen;

Fig. 5g

illustriert schematisch eine möglich Genauigkeit des ML-Modells (mashine lerning, Maschinen Lernen) gemäß Ausführungsbeispielen;

Fig. 5h

zeigt schematisch Garstellung eines Markow-Entscheidungsproblems für fünf Kochzustände mit beispielhalft gewählten Übergangswahrscheinlichkeiten (Ziel ist es, die optimale Aktion (hier: Leistungsstufe 6) zu bestimmen, um den ZIEL-Kochzustand (hier: Kochen) zu erreichen) zur Erläuterung von Ausführungsbeispielen;

Fig. 5i

zeigt exemplarisch eine Hardware-Vorauswahl und Systemaufbau des Sensor mit Syntiant NDP120 zur Erläuterung von Ausführungsbeispielen.

Embodiments of the present invention are explained with reference to the accompanying drawings.

Fig. 1A-1B

schematic representations of a cooking appliance to explain the functionality in embodiments;

Fig. 1C

schematic diagrams illustrating power control in embodiments;

Fig. 1D to 1G

schematic representations to illustrate the control of the temperature according to embodiments;

Fig. 2A

a schematic block diagram to illustrate the integration of means for sound detection in a control system for the cooking appliance according to embodiments;

Fig. 2B

a schematic illustration to explain different features in the analysis according to embodiments;

Fig. 2C

a schematic overview to explain possible implementation variants for means for sound detection according to embodiments;

Fig. 2D

a schematic representation ( Fig. 2D left, 2Dright) to explain possible positioning of the sound detection means in cooking appliances;

Fig. 2E and 2F

schematic diagrams for explaining and detecting noises that can be filtered according to embodiments;

Fig. 2G and 2H

schematic illustrations to explain the filtering of noise according to embodiments;

Fig. 3A and 3B

schematic representations to explain the classification of cooking states according to embodiments;

Fig. 3C-3G

schematic representations to explain applications according to embodiments;

Fig.4

a schematic representation of a cooking appliance with means for sound detection and a processor according to a basic embodiment of the cooking appliance;

Fig. 5a

schematically illustrates the design of the cooking appliance or sensor as a cyber-physical system to illustrate embodiments;

Fig. 5b

schematically illustrates the classification of standard cooking methods according to temperature (at normal pressure) and audibility to explain exemplary embodiments;

Fig. 5c

schematically illustrates characteristic features of cooking noises, shown as a Mel spectrogram to explain embodiments;

Fig. 5d

schematically illustrates an exemplary structure and application of a CNN using the example of four cooking states according to embodiments;

Fig. 5e

schematically illustrates a test setup and hardware for data acquisition as part of the feasibility study to explain implementation examples;

Fig. 5f

schematically illustrates selected Mel spectrograms (200x200) for each cooking state from the training data set (observation period within a spectrogram is 2.14 s on the ordinate, the number of Mel filter bands is 200, with a bandwidth of 200 Hz to 20,000 Hz) to explain embodiments;

Fig. 5g

schematically illustrates a possible accuracy of the ML model (machine learning) according to embodiments;

Fig. 5h

shows a schematic representation of a Markov decision problem for five cooking states with exemplary transition probabilities (the aim is to determine the optimal action (here: performance level 6) to reach the TARGET cooking state (here: cooking)) to explain embodiments;

Fig. 5i

shows an example of a hardware preselection and system structure of the sensor with Syntiant NDP120 to explain implementation examples.

Before embodiments of the present invention are explained below with reference to the accompanying drawings, it should be pointed out that elements with the same effect are provided with the same reference numerals, so that the description of them is applicable to one another or interchangeable.

Fig. 1A shows a cooking device 10, which here can have, for example, four heat sources 12a-d. Each of these heat sources 12a-d can be individually controlled, for example, via a control 20, e.g. via a rotary knob 20k, or can also be changed in terms of size.

The visible top of a hob 10 usually consists of special glass ceramic surfaces with several cooking zones 12a-12d (heating area), which mark the area of the induction coils. The four rotary knobs 12k, marked with numbers, give the cook the option of setting a target value for the heating power for each cooking zone. In addition to radiant heating, hobs with induction technology are also offered on the household appliance market. The main difference to radiant heating (infrared heating) is that the heat is generated directly in the base of the cookware (not shown). Accordingly, the cooking zone is only heated by the heat from the cookware and the surface temperature of the glass ceramic plate remains low. In addition, an induction hob can heat food more quickly than another type of hob. With reference to Fig. 1B This type of hob is explained.

Fig. 1B shows a hob 10 with a cooking utensil 15. The cooking appliance 10 has a cooking zone 12a that can be operated using induction. The cooking zone 12a comprises several induction coils 12i that are driven by a converter 12u. The converter receives its electrical energy from the mains and is controlled by the controller 20.

1. 1) Die Frequenz des Wechselstroms aus dem Stromnetz wird mithilfe des Umrichters 12u (meist einem Royer-Oszillator) von 50 Hz auf 20-60 kHz erhöht.
2. 2) Der hochfrequente Wechselstrom wird in die Induktionsspulen 12i gleitet, wodurch ein magnetisches Wechselfeld 12w für die Energieübertragung entsteht.
3. 3) Das magnetische Wechselfeld wird im Pfannenboden des Kochgefäßes 15 gebündelt und induziert dort einen elektrischen Wirbelstrom.
4. 4) Da das ferromagnetische Material im Pfannenboden einen deutlich höheren ohmschen Widerstand als die Induktionsspule aufweist, wird dort (unter anderem aufgrund des Skin-Effekts und Ummagnetisierungsverlusten) der größte Teil des Wirbelstroms in Wärme umgewandelt.

The heat development in the induction hob basically occurs in four steps:

1. 1) The frequency of the alternating current from the power grid is increased from 50 Hz to 20-60 kHz using the 12u converter (usually a Royer oscillator).
2. 2) The high frequency alternating current is fed into the induction coils 12i, creating an alternating magnetic field 12w for energy transfer.
3. 3) The alternating magnetic field is concentrated in the bottom of the pan of the cooking vessel 15 and induces an electrical eddy current there.
4. 4) Since the ferromagnetic material in the pan bottom has a significantly higher ohmic resistance than the induction coil, most of the eddy current is converted into heat there (due to the skin effect and remagnetization losses, among other things).

The excitation of the induction coils 12i is carried out with a corresponding excitation frequency as Fig. 1C shows. Here the available electrical power P _eff is plotted over time. Fig. 1C shows three diagrams with three different output powers. To control the power, the control 20 can use modulation. Starting from an exemplary Royer oscillator (with a fixed frequency of the coil current), the effective electrical power can be controlled using the programmed modulation of the pulse duration (PDM) - i.e. by simply switching the oscillator on and off. Depending on the setting duration and the pauses, an effective active power P _eff is thus obtained. A long setting duration and short pause lead to the food being cooked heating up quickly and vice versa.

In addition to the classic method of controlling the power, high-quality hobs are also offered with temperature control. For temperature control in the sense of a closed-loop system 20l, either the temperature sensor under the glass-ceramic hob is used or special cooking and frying sensors with radio transmission are used (with the advantage that the temperature of the food being cooked can be measured directly). In connection with the Royer oscillator (fixed frequency of the coil current), a so-called two-point controller 20z is usually used via the control system 20r. With this two-point controller, the excitation frequency of the induction coil can be adjusted, as shown by the curve f20. The closed-loop system 20l is It should be noted that, starting from the desired power w(t) and a temperature signal x(t), the two-point controller 20z is controlled with the combined signal e(t) in order to apply the excitation frequency t20 or y(t) to the induction coils.

If a converter is used instead of a Royer oscillator, which enables a continuous variation of the excitation frequency, a continuously operating controller 20pit can be used for the control system 20r of the modified loop 20l'. Accordingly, the controller 20pit does not adjust the duty cycle of the oscillator, but rather the control frequency of the coil variably, as shown by f20' (associated with y(t)).

In the examples explained above from the Fig. 1D/1E and 1F/1G the control is carried out by means of the controller 20z or 20pit using a temperature sensor 20ts. As already explained above, this temperature sensor can be integrated into the glass ceramic hob and thus measures the pot temperature or can be present in the hob. Either the ergonomics or the indirect measurement of the medium temperature are problematic. The only boiling states of the medium temperature can vary depending on the medium or other environmental conditions, such as the altitude. The boiling point varies both depending on the altitude and depending on the medium. Regardless of the medium, however, boiling can be detected based on the sound emission of the medium in the fluid. This effect is subsequently used by the cooking device 50. The cooking device 50 comprises, for example, a heating element, such as an induction heating element 52, e.g. an induction coil, as explained above. The cooking device 50 also comprises means for sound detection 54, such as a microphone or a structure-borne sound sensor. The means for sound detection 54 are designed to detect the acoustic signal 56 originating from the medium 58 in the cooking utensil 59. The acoustic signal is then converted, for example, into a digital acoustic signal and evaluated by the processor 55. The evaluation can be carried out according to embodiments according to one or more features in order to determine the cooking state of the medium 58 based on these one or more features. For example, boiling information can be recognized based on the noise. As explained above, an acoustic signal is produced by the medium during the boiling process, so that this information can be generated, for example, by means of pattern recognition. It has also been recognized that temperature information or other information, such as an evaporation or combustion state of the medium, particularly in the case of fat, can also be recognized.

According to optional embodiments, the processor 55 outputs this status information to the controller 20. Based on this, the controller 20 can control the heating power for the heating element 52, as described above in connection with the control examples from Fig. 1D , and 1F was explained.

The following refers to Fig. 2a a variant is explained in which the acoustic sensor 54 and the processor 55 are integrated into an external device, such as a retrofit device 53. The functionality is essentially the same, with the processor 55 then outputting the generated information and/or features as information to a programming interface of a standard hob. Here, the programming interface API is provided with the reference symbol 20a. Like the programming interface, according to further embodiments, not only the generated information, e.g. in the form of control data, can be output to the API interface, but also additional information, such as a rotary knob position, a pan detection and/or a temperature can be tapped from the interface 20a by the device 53 and included in the control. Based on this, the conventional control can then take place using the controller 20z and the control path 20r. At this point, it should be noted that the controller is of course not limited to a two-point control; a PID controller would also be conceivable.

The only sensor signals that can be included in the monitoring implementation examples are in Fig. 2B illustrated. In this embodiment, it is assumed that the acoustic sensor 54 can be integrated into the ceramic hob. The ceramic hob optionally includes a pan detection 61 and a temperature sensor 62. Based on this, the four input data rotary knob position, pan detection, temperature and the acoustic signal are then available, so that using these signals the excitation frequency for the controller 20, i.e. the power control, can be adjusted. This control can, for example, be carried out separately for each cooking zone, i.e. x times. The rotary knob position enters the setpoint, whereby the excitation frequency is then selected taking temperature and acoustic signal into account so that a suitable cooking state is achieved. This is recognized and regulated via the acoustic signal and/or the additional signals. When teaching the model, the processor (not shown) creates an ML model by drawing conclusions about the dynamic system of controller and controlled system (glass ceramic plate and pot) and teaches this accordingly. Knowledge about the pot contents can also be obtained from the amount of energy supplied and the associated temperature change (temperature sensor 62). The pot content is therefore a corresponding feature associated with the cooking state, since both the acoustic characteristics and the cooking state itself can vary depending on the medium or pot content.

According to embodiments, 54 different variants are considered as means for sound detection. According to a first variant, as for example in connection with Fig.3 As explained, airborne sound can be determined. According to one embodiment, a MEMS microphone could be used for this. According to another variant, the determination of structure-borne sound would also be conceivable.

A structure-borne sound sensor (structure-borne sound microphone) is an electroacoustic transducer for measuring structure-borne sound, whereby the measurement mainly relates to the investigation of vibrating surfaces. The most important parameters are the deflection, the vibration speed and the vibration acceleration. Accordingly, four types of structure-borne sound sensors are placed, for example, as in Fig. 2C These include piezoelectric sensors, MEMS sensors, strain gauges, magnetic-inductive sensors. All are used to record acceleration.

As in Fig. 2B As shown, the sound sensor, such as the microphone or the structure-borne sound sensor, can be positioned at different positions of the cooking surface 10 with the multiple cooking zones. In the right Fig. 2B On the right, four sound sensors are provided, each assigned to the cooking zones 12a-d. Alternatively, the sound sensors can also be provided between the cooking zones, as in the left Fig. 2B illustrated on the left. In this case, a 1:1 assignment of sound sensor and cooking zone is not possible.

According to embodiments, a microphone array can also be used as a microphone. This offers several advantages, namely that several cooking zones 12a-d can be monitored using one array, for example using beamforming. According to embodiments, background noise can also be suppressed. The use of an array in particular makes it possible to detect existing background noise and then take it into account or filter it out (before further processing of the signals). According to embodiments, the microphones can be attached directly to the hob or in the cooking environment (e.g. extractor hood, kitchen ceiling, etc.).

1. 1) Ein Royer-Oszillator erzeugt Störgeräusche mit einer konstanten Frequenz (vgl. Fig. 2E).
2. 2) Ein Umrichter mit variabler einstellbarer Frequenz erzeugt ein Störgeräusch mit variierender Frequenz (vgl. Fig. 2F).

The active induction coils generate noise that may need to be filtered. Noise occurs in different ways depending on the converter:

1. 1) A Royer oscillator generates noise with a constant frequency (cf. Fig. 2E ).
2. 2) An inverter with variable adjustable frequency generates a noise with varying frequency (cf. Fig. 2F ).

An analysis of the noise is shown, for example, between stimulation pauses, such as Fig. 2G in area A.

According to embodiments, the processor has a corresponding filter to filter out such noise, which can be calculated depending on the choice of converter. A simple possibility would be to use a low-pass filter to filter the audio signal, as in Fig. 2H is shown by the reference number 55t. According to embodiments, the low-pass filter also receives information such as the current excitation frequency in order to optimally filter the audio signal.

In order to recognize one or more characteristics of a cooking state based on the acoustic signal, the processor is trained in advance or supplied with trained data. Using machine learning, the different signals, e.g. knob position, pot detection, temperature, excitation frequency and, above all, the acoustic signal, can be combined and used to determine a pattern for different cooking states. An ML model trained in this way (SVM, CNN, TCNN, RNN, LSTM, Transformer) can then be used to subsequently monitor and automate the cooking process. According to a basic embodiment, the acoustic emission from cooking processes or other events on the stove or similar kitchen appliances is recorded using structure-borne and/or airborne sound (audible frequencies up to 20,000 Hz). These acoustic emissions are then assigned a cooking state that can be recognized at a later point in time. According to further embodiments, sensor data is processed by intelligent signal analysis or preprocessing, e.g. time/frequency transformation or feature extraction, in combination with machine learning methods, e.g. B. by neural networks, continuously or at fixed time intervals. The airborne and/or structure-borne sound data are combined with other Sensor data is merged. The analysis can be carried out locally directly on the device or sensor. Alternatively, cloud evaluation is also possible. An ML model (machine learning model) trained in this way then enables various one or more previously defined states to be recognized, e.g. water boiling over, and this, for example, independently of the previous signal curve. Instead of fixed events or in addition to fixed events, continuous events, such as a "sizzling intensity" of fat, can also be determined.

Such a device can be integrated directly into the kitchen appliance, as is the case, for example, with Fig.4 explained or can also be intended as a retrofit element, such as Fig. 2A has shown.

The integrated variant or the retrofit variant both offer the possibility of providing feedback to the kitchen appliance based on the results of the analysis in order to regulate its parameters, in accordance with other embodiments. Additionally or alternatively, information can also be issued to a user, e.g. via smartphone, to inform the user of the current status and, if necessary, to point out problems.

Such problems can be interpreted as boiling conditions. In connection with Fig. 3B Three cooking states are shown as examples. Fig. 3B Based on a large number of linked sensor information (cf. Fig. 3a ), which are weighted according to their relevance, for example, three resulting recognized states "water is boiling over", "fat is sizzling too much" or "everything is OK". These are example classes for safety functions. Based on each state, a hint can then be given or the hob power can be controlled directly. If water is boiling over or fat is sizzling too much, for example, the hob power (or the temperature setpoint) can be reduced. If everything is OK, the setpoint value from the rotary knob is used. According to the predictions of the ML model, this can be done via the manufacturer's API in order to make a temperature adjustment. Both classification and regression models can be used. It can also be checked whether a classification is appropriate or whether the model (e.g. prediction of a temperature value) should be improved. The user can then provide feedback via the API, which is taken into account (see below).

According to the embodiments, the safety function is to be understood as a basic function. In particular, kitchen fires often occur as a result of burning fat. The condition of heavily sizzling fat can thus lead to a reduction in the heating output and can advantageously be detected very well using structure-borne sound or airborne sound. Essentially the same applies to boiling over, as the boiling over process can be identified very well using the acoustic model, even if the potential damage effects are less, since no fat fire results here. Nevertheless, a pot boiling over is a safety-critical issue, so that this detected condition can also be assigned to a safety function. According to the embodiments, it would also be conceivable to deactivate this safety function in order to actively introduce very high power into the medium, e.g. when deep-frying. The self-learning algorithm can also be further trained using user feedback if, for example, it detects a pot boiling over or detects a fat temperature that is too high even though this has not yet been reached. In this respect, user feedback regarding status information can also enable training during operation.

At this point, it should be noted that other cooking appliances, such as a kettle, could also potentially be used. A kettle has an integrated heating element and can be monitored acoustically. The kettle can then be switched off based on the acoustic signal.

In case of cooking water can be boiled using conventional cookware (see Fig.3D ) or by means of a stove-operated kettle (cf. Fig. 3C ) the kettle functionality "boiling once" can also be enabled. The hob is then switched off as soon as the water has boiled. According to the examples, it would of course also be conceivable that voice control is possible via such functions, since the safety-relevant aspects can be secured by acoustic monitoring.

Fig.3D shows a possible variant of how further information can be transmitted, e.g. from the cookware to the control system. For example, an RFID tag can be provided as a sticker on the cookware, which outputs information about the functionality to the cooking device via RFID. In this respect, the cooking device can be used according to embodiments have an RFID reader. According to embodiments, a dynamic model or special function can be stored in an RFID tag and transmitted to the intelligent control of the hob. An interesting option would be an RFID tag that gives old pots or similar new functions. For example, "I am a kettle" in order to carry out the kettle functionality explained above.

This means that, in addition to or as an alternative to the safety functionalities, a convenience functionality, such as a cooking mode assigned to a cooking task, can also be enabled.

Fig. 3E shows a controller 20' with a processor 55, which receives the acoustic signal from the sensor 54. The controller can, according to embodiments, output information to the user, for example via Bluetooth or other radio communication means. The user can, for example, wear headphones shown here for voice output or a smartwatch for visual display or alarm via vibration signal. Furthermore, the sound information from the cooking area / cooking status can also be converted into representative vibration patterns and then transmitted to a smartwatch.

Another variant of a possible control variant or convenience functionality could be the situation-dependent regulation of the power. For example, according to embodiments, the cooking medium in the cookware can be kept constant in a first state, which is detected, for example, upon confirmation by the user. If the user or the cook is satisfied with the current cooking or sizzling intensity, he or she can communicate this to the control system - similar to cruise control in a car - using a hold button. The control system then ensures that the cooking or sizzling intensity is maintained accordingly. The pasta water does not boil over. There is therefore no need to repeatedly readjust the heating power using the rotary knob. Such a function is ideal for gently simmering, warming up or keeping food warm. In addition, a liquid often starts to boil more strongly immediately if you want to use a lid to save energy.

According to embodiments, the control can differentiate between the state with lid and without lid based on the acoustic signal and then adjust the regulation accordingly. Depending on the state of lid or without lid, Both the acoustic signal and the behavior of the cooking medium, which the control system is designed to take into account, must be taken into account. A specially optimized warm-up program would also be conceivable, as will be explained below.

A warm-up program or a cooking program in general could, for example, carry out an automated regulation of the heating power according to a given recipe. Content stored in the recipe can, for example, be the duration of individual heating levels or the duration of individual cooking states. For example, it can be defined that cooking state A is maintained for a certain period of time, while cooking state B is maintained for a further certain period of time. Warm-up programs could, for example, be designed in collaboration with manufacturers of ready meals. Using a QR code on the packaging, the intelligent control can then execute a cooking program optimized by the manufacturer (amount of water + powder in the pot, scan QR code with hob/smartphone, click start, if necessary beep to stir, etc.). The same principle can be applied in cookbooks. Furthermore, different programs are also made possible in this way, for example for warming sausages such as white sausages, according to a given recipe.

Referring to Fig. 3F The determination of the data basis for the temperature prediction is explained using a neural network that uses the structure-borne sound to predict the temperature of the fat or the medium in the pan or in the cookware. For training purposes, the temperature of the medium to be heated is determined using an infrared thermometer and at the same time the currently prevailing structure-borne sound is recorded using a sound sensor, e.g. a piezo pickup or a microphone. These two pieces of information can be correlated. The relationships can be learned using an ML model (regression model). Information about the medium to be heated can also be read in as a further feature to be correlated. The background is that, as the table in Fig. 3F shows that different media have different smoke points because they can be heated to different maximum temperatures.

Fig. 3G shows an extract from a CSV marker for, for example, three different cooking intensities KS0 stove, KS1 light cooking, KS2 strong cooking, KS3 boiling over.

In the following, a preferred embodiment is described together with the functional principle using Fig. 5A-I explained in detail.

One example creates an intelligent cooking sensor that can reliably detect cooking conditions and regulate them automatically. It consists, for example, of an artificial auditory perception model and a new type of power controller. Together with an induction hob, it thus forms an intelligent cyber-physical overall system. It can perceive and understand the cooking environment and change it in a targeted manner through actions - such as adjusting the heating output. With the ability to act autonomously, it forms the foundation for the smart hob of the future and the basis for user-friendly cooking assistance systems.

Due to the intended system structure, the sensor can be easily and quickly integrated into conventional hob systems.

The functional principle of the sensor is explained by Fig. 5a explained. Fig. 5a illustrates the design of the sensor as a cyber-physical system. Here, the three stages of perception model, system state and system controller are shown depending on the target and actual cooking state, taking the pot dynamics into account.

The sound pressure waves perceived as environmental noise are recorded by inexpensive MEMS microphones, pre-processed in a digital signal processor and converted into a spectrogram. This contains the noise information and serves as input for the cooking state detection in the artificial auditory perception model. The output is temporarily stored as the actual cooking state. In addition to the recorded audio signals AS, the information from the hob control is included in the perception (sensor fusion). This includes: indirect temperature measurements, pot detection and current heating output (see. Fig. 3a ).

The system state M forms the input variable for the power controller as a state vector. In addition to the perceived ACTUAL cooking state, the system state Z contains the desired TARGET cooking state and an internally calculated parameter for the pot dynamics. The latter characterizes the thermodynamic behavior of the entire system consisting of the hob, cookware and liquid. The system state Z can also be expanded to include specifications for cooking modes (e.g. energy saving mode).

The state-based power controller R calculates the currently required heating power based on the system status and sends either a command to the hob control or a warning message to the person cooking. In this way, the TARGET cooking state can be achieved and maintained autonomously and energy efficiently.

The concept of classifying cooking states is described below.

Audible cooking methods and useful cooking states are in Fig. 5b shown: Cooking or cooking is generally understood to mean the preparation of food using a heat source. A basic distinction is made between moist cooking methods (cooking in the true sense of the word) and dry cooking methods (frying, deep-frying and sweating). The main difference between the two methods of preparation is the medium used to transfer the heat to the food. For example, liquids such as water, broth, wine and milk are used for cooking; for frying, deep-frying and sweating, heating is done using fats and oils [8]. The choice of cooking method has a direct impact on the taste, tolerability and nutritional content of the prepared food - and often also on health [8]. An overview of the cooking methods regularly used in private households can be found in Fig. 5b Cooking methods that can be heard acoustically are shaded.

• Kochzustände sind meist über einen langen Zeitraum konstant zu halten
• Es sind weniger direkte Interaktionen erforderlich als z. B. beim Braten
• Sie machen den größeren Teil aller Garmethoden aus und sind allgemein gesünder
• Das Potential zum Energiesparen ist größer
• Der Mehrwert für die Endbenutzerinnen ist insgesamt größer

For the demonstrator to be developed to acquire customers and for a first minimally practicable product (MVP) in the sense of the Lean Startup method, the moist cooking methods are best suited. Because:

• Cooking conditions usually have to be kept constant over a long period of time
• Less direct interaction is required than, for example, when frying
• They make up the majority of all cooking methods and are generally healthier
• The potential for energy saving is greater
• The overall added value for end users is greater

As part of a feasibility study for the artificial auditory perception model, the following cooking state classes were used: stove off, implosions, simmering, boiling, boiling, vigorous boiling and boiling over. This allows almost all moist cooking methods to be monitored and controlled based on their specific cooking noises. The state classes can be used as part of customer discovery. In this context, safety-relevant functions for roasting and deep-frying as well as roasting assistants in connection with sizzling intensities can also be implemented.

According to examples, hearing can be used with artificial intelligence: Noises are generally auditory sensations that are not directly perceived as a sound, tone, mixture of tones, harmony or bang. The cause of noises are vibration processes mediated by elastic bodies (e.g. a cooking pot) that spread as airborne sound in the room [9]. In contrast to a whistling sound, a cooking noise does not have an exactly definable pitch, i.e. no dominant frequency (see Fig. 5c ). Nevertheless, different frequency ranges are represented to varying degrees and thus give cooking noise an individual character, which can be seen in the Mel spectrogram MS shown in Fig. 5c can be clearly seen. Note: A spectrogram is the graphical representation of the temporal progression of the frequency spectrum of a signal (e.g. an audio signal) [10]. The time signal is converted into the time-frequency domain using the short-time Fourier transform (STFT) [11]. In a Mel spectrogram, the frequencies are shown on the Y-axis in the Mel scale [12]. In addition, striking temporal patterns also emerge. In addition or as an alternative, the temporal progression could also be learned instead of frequency ranges.

Based on such characteristic features, different sounds can be classified by both the human brain and an ML model and thus assigned to a known cooking state. It should be noted that the basic characteristics of the cooking sounds are retained despite superimposed noise S. In addition, microphones are not limited to the human hearing range.

In order to automatically analyze the acoustic signals, machine learning (ML) methods are used, e.g. in the form of deep neural networks, according to the embodiments. In supervised learning, the neural network learns the relationships between cooking noise and cooking state using training data. After successful training, the neural network (ML or Kl model) can be used to classify cooking noises (see Fig. 5d ).

1. 1. Datenerhebung: Im Rahmen der Datenerhebung wurden Kochgeräusche mit Hilfe eines Mikrofons aufgenommen und in einem Trainingsdatensatz zusammengefasst. Ein einzelnes Datensatzelement besteht dabei aus einer Audiodatei (Kochgeräusch) sowie dem zugehörigen Klassenlabel (Kochzustand). Bei der Aufnahme der Trainingsdaten kam das vorab entwickelte Aufnahme und Label-Programm "KitchenGuard Audio Recorder" zum Einsatz, welches eine schnelle und fehlerfreie Datenerhebung erlaubt. Auf dieser Basis konnte ein erster Trainingsdatensatz für die sieben definierten Kochzustände erstellt werden. Dieser beinhaltet 748 Aufnahmen mit einer Gesamtdauer von etwa 12,5 Stunden. Der Versuchsaufbau und die verwendete Hardware sind in Fig. 5e dargestellt.
2. 2. Vorverarbeitung, Merkmalsextraktion: Aus den aufgezeichneten Audiodaten wurden im Rahmen der Vorverarbeitung gelabelte Mel-Spektrogramme extrahiert (siehe Fig. 5f). Aufgrund der moderaten Größe des Trainingsdatensatzes, konnten mit einem Betrachtungszeitraum von 2,14 s und folglich 17.574 Mel-Spektrogrammen die besten Ergebnisse erzielt werden. Um den realen Betrieb mit Störgeräuschen besser abzubilden, wurden die Trainingsbeispiele im Rahmen einer Datenaugmentierung (engl. Data Augmentation) erweitert und diversifiziert, wie Fig. 5f darstellt.
3. 3. Modellauswahl und Training: In der Machbarkeitsstudie wurde u. a. das von Google LLC entwickelte CNN Mobilenetv2_050 getestet [Q]. Dieses CNN ist speziell für den Einsatz auf mobilen Geräten optimiert und kann zuverlässig auf Mikrocontrollern und kleinen Kl-Prozessoren (microNPUs) betrieben werden. Das CNN wurde daraufhin mit dem erstellten Trainingsdatensatz trainiert.
4. 4. Bewertung: Um die Leistungsfähigkeit eines Klassifizierungsmodells zu bewerten, wird vornehmlich die Genauigkeit (engl. Accuracy) in Bezug auf die Vorhersage von unbekannten Testdaten als erstes Bewertungskriterium herangezogen. Die Genauigkeit des ML-Modells beschreibt das Verhältnis zwischen den richtigen Vorhersagen und allen getroffenen Vorhersagen. [14]. Die Ergebnisse sind in Fig. 5g dargestellt.

A feasibility study for classifying cooking sounds was conducted as follows. The functionality of the ML-based auditory perception system was demonstrated.

1. 1. Data collection: As part of the data collection, cooking noises were recorded using a microphone and combined in a training data set. A single data set element consists of an audio file (cooking noise) and the associated class label (cooking state). The previously developed recording and labeling program "KitchenGuard Audio Recorder" was used to record the training data, which allows for fast and error-free data collection. On this basis, a first training data set for the seven defined cooking states could be created. This contains 748 recordings with a total duration of around 12.5 hours. The test setup and the hardware used are described in Fig. 5e shown.
2. 2. Preprocessing, feature extraction: Labeled Mel spectrograms were extracted from the recorded audio data during preprocessing (see Fig. 5f ). Due to the moderate size of the training data set, the best results were achieved with a viewing period of 2.14 s and consequently 17,574 Mel spectrograms. In order to better reflect real operation with noise, the training examples were expanded and diversified as part of data augmentation, as Fig. 5f represents.
3. 3. Model selection and training: The feasibility study tested, among other things, the CNN Mobilenetv2_050 developed by Google LLC [Q]. This CNN is specially optimized for use on mobile devices and can be reliably operated on microcontrollers and small microprocessors (microNPUs). The CNN was then trained using the training data set created.
4. 4. Evaluation: To evaluate the performance of a classification model, the accuracy in predicting unknown test data is used as the first evaluation criterion. The accuracy of the ML model describes the ratio between the correct predictions and all predictions made. [14]. The results are shown in Fig. 5g shown.

It can be seen that even with little training data, comparatively good accuracies were achieved. The lower accuracies of the simmering, boiling and strong boiling state classes are due, among other things, to the fact that the noises caused in these states are very similar and are therefore more easily confused. More training examples are required to learn subtle differences. However, the results presented are sufficient to demonstrate feasibility.

Finally, it should be noted that there are many other ML methods for noise analysis. A number of these methods have been summarized by the experts in a toolkit for acoustic monitoring (amo for short) [15]. These are also tested and evaluated for the sensor. (Note: Test data is data that was not shown to the ML model during training and is therefore unknown. They therefore represent real-life use (e.g. in the end customer's kitchen). Accordingly, a separate test data set was recorded for the evaluation of the MLK model, in which the recording day, pot manufacturer and pot contents differ from the training data. In addition, existing noise represents real operation.

According to embodiments, a power controller can be used to calculate the optimal heating power: The sensor's power controller has the task of selecting the best possible power level based on the current system state. The current system state includes the perceived ACTUAL cooking state, the desired TARGET cooking state and the calculated pot dynamics.

In contrast to a control based on temperature measurements or vibration intensities, no classic concepts of control engineering can be applied here. In addition, since there is no perception model that can determine the actual cooking state with 100 percent accuracy and the pot dynamics can also change (e.g. when the user adds cold water), this is a non-deterministic decision-making process; the consequence of a selected action cannot therefore be predicted exactly.

The optimal performance level can still be determined on the basis of the Markov Decision Process (MDP). For example, using a learned strategy (policy) or using statistically recorded, action-dependent transition probabilities. There are methods for this that fall into the areas of dynamic programming, reinforcement learning and imitation learning, among others. By selecting the appropriate boundary conditions, safety-critical conditions can be safely avoided (e.g. boiling over). A simplified example with action-dependent transition probabilities is shown in Fig. 5h shown.

The planned action (the power level that is most likely to lead to the desired TARGET cooking state) is finally transmitted to the hob control.

• Privatsphäre und Datenschutz: Private Daten bleiben in der Küche (z. B. Gespräche).
• Sicherheit: ML-Funktionen stehen ohne aktive Internetverbindung zur Verfügung.
• Geschwindigkeit: Kochgeräusche können in Echtzeit im Sensor ausgewertet werden.
• Energieeffizienz: Die Leistungsaufnahme beträgt i. Allg. weniger als 500 µW.
• Preiswert: Zusätzliche Chips für Ethernet oder WLAN werden nicht benötigt und es entstehen nahezu keine laufenden Betriebskosten. [17]

Durch die steigende Zugänglichkeit der kostengünstigen, MLK-fähigen Mikrocontroller ist der Zeitpunkt für die Umsetzung des Sensors ideal.Regarding the possible hardware for the sensor: The application of a trained ML model is called inference (model inference). In contrast to model training, this requires significantly less computing power and RAM. For this reason, the auditory perception model of the sensor can also be operated on a low-cost microcontroller (keyword: EndpointAl13, TinyML14). This allows the cooking noises to be classified directly in the hob and in real time (inference time of less than 100 ms) [16]. In contrast to classic, cloud-based inference operation, the following advantages arise:

• Privacy and data protection: Private data remains in the kitchen (e.g. conversations).
• Security: ML functions are available without an active internet connection.
• Speed: Cooking noises can be evaluated in real time in the sensor.
• Energy efficiency: The power consumption is generally less than 500 µW.
• Inexpensive: Additional chips for Ethernet or WLAN are not required and there are almost no ongoing operating costs. [17]

With the increasing availability of low-cost, MLK-capable microcontrollers, the timing for implementing the sensor is ideal.

This in Fig. 5i The electronic hardware concept presented represents an initial selection of hardware components based on current findings and assumptions. This hardware structure allows four cooking zones to be monitored and controlled simultaneously. Accordingly, four high-quality MEMS microphones15 are provided for audio signal recording. The selection, positioning and alignment of these microphones, the so-called microphone concept, is to be developed as part of the EXIST start-up grant.

On the vision of a user-friendly and safe cooking assistant: In order to meet young people's need for technology and luxury and at the same time provide intuitive operation that is accessible to technology-averse end customers, the assistant is intended to be a smart cooking assistant. Interactive cooking apps are intended to help people cook healthy food in a fun and stress-free way, discover new recipes and test new energy-saving and automated cooking methods (e.g. preparing eggs using the Ogi method). There is also the possibility of giving manufacturers of ready-made products, baby food and cookbook publishers the opportunity to provide interactive apps themselves and thus ensure optimal preparation of their products. Other considerations include: a connection to smart refrigerators for recipe recommendations, specially developed, sound-emitting cookware or the possibility of live transmission of cooking sounds and visual information on smartphones or VR glasses. The aim is to make cooking as attractive, interactive and relaxing as possible while providing the necessary safety.

Different embodiments and optional features for embodiments are explained below. Depending on the embodiments, it is possible to train the system, i.e. the cooking appliance or cooking accessories, or to continue to learn it. Federated training (English: Federated Learning), distributed across multiple users, would also be conceivable. This would then expand the database across multiple devices. The user data can only be partially annotated or not annotated at all and contribute to improving the model through partially supervised learning.

According to embodiments, the processor is designed to carry out an adaptation phase in which the device or the recognition algorithm of the device adapts to the current cooking device or cookware.

According to the exemplary embodiments, the evaluation and, if necessary, further learning is carried out using privacy enhancement technologies (homomorphic encryption, differential privacy, secure multi-party computation, or similar) in order to avoid the unwanted disclosure of sensitive information or to enable conclusions to be drawn about the participants, attributes of the participants or other sensitive information.

According to embodiments, the processor may be configured to remove the speech as personal information, possibly by tools for filtering out speech information, such as voice activity detection.

According to a further embodiment, it would be conceivable that a hotplate or cooking zone on which the pot is located is recognized and assigned. For example, the location of noises can be used for this purpose or information can be obtained from the cooking appliance, e.g. about an activated cooking zone.

In all embodiments, it is crucial that air and/or structure-borne sound sensors are used to analyze the cooking processes.

In addition to the hob/stove, possible applications include hotplates, kitchen appliances with cooking functions such as Thermomix, kettles or even ovens.

At this point, it should be noted that, according to embodiments, the means for sound detection can also comprise ultrasonic sensors. In addition, it would also be conceivable that a camera sensor system is used in addition to the sound sensor system.

Although some aspects have been described in the context of a device, it is to be understood that these aspects also represent a description of the corresponding method, so that a block or component of a device can also be understood as a corresponding method step or as a feature of a method step. Analogously, aspects described in the context of or as a method step also represent a description of a corresponding block or detail or feature of a corresponding device. Some or all of the method steps can be performed by a hardware apparatus (or using a hardware apparatus), such as a microprocessor, a programmable computer, or an electronic circuit. In some embodiments, some or more of the key method steps can be performed by such an apparatus.

A signal encoded according to the invention, such as an audio signal or a video signal or a transport stream signal, can be stored on a digital storage medium or can be transmitted on a transmission medium such as a wireless transmission medium or a wired transmission medium, e.g. the Internet.

The encoded audio signal according to the invention can be stored on a digital storage medium, or can be transmitted on a transmission medium, such as a wireless transmission medium or a wired transmission medium, such as the Internet.

Depending on particular implementation requirements, embodiments of the invention may be implemented in hardware or in software. The implementation may be performed using a digital storage medium, for example a floppy disk, a DVD, a Blu-ray Disc, a CD, a ROM, a PROM, an EPROM, an EEPROM or a FLASH memory, a hard disk or other magnetic or optical storage on which electronically readable control signals are stored that can or do interact with a programmable computer system in such a way that the respective method is carried out. Therefore, the digital storage medium may be computer readable.

Some embodiments according to the invention thus comprise a data carrier having electronically readable control signals capable of interacting with a programmable computer system such that one of the methods described herein is carried out.

In general, embodiments of the present invention may be implemented as a computer program product having a program code, wherein the program code is operable to perform one of the methods when the computer program product is run on a computer.

The program code can, for example, also be stored on a machine-readable medium.

Other embodiments include the computer program for carrying out one of the methods described herein, wherein the computer program is stored on a machine-readable medium. In other words, an embodiment of the method according to the invention is thus a computer program that has a program code for carrying out one of the methods described herein when the computer program runs on a computer.

A further embodiment of the methods according to the invention is thus a data carrier (or a digital storage medium or a computer-readable medium) on which the computer program for carrying out one of the methods described herein is recorded.

A further embodiment of the method according to the invention is thus a data stream or a sequence of signals which represents the computer program for carrying out one of the methods described herein. The data stream or the sequence of signals can be configured, for example, to be transferred via a data communication connection, for example via the Internet.

A further embodiment comprises a processing device, for example a computer or a programmable logic device, which is configured or adapted to carry out one of the methods described herein.

A further embodiment comprises a computer on which the computer program for carrying out one of the methods described herein is installed.

A further embodiment according to the invention comprises a device or a system which is designed to transmit a computer program for carrying out at least one of the methods described herein to a recipient. The transmission can be carried out electronically or optically, for example. The recipient can be, for example, a computer, a mobile device, a storage device or a similar device. The device or system can, for example, comprise a file server for transmitting the computer program to the recipient.

In some embodiments, a programmable logic device (e.g., a field programmable gate array, an FPGA) may be used to perform some or all of the functionality of the methods described herein. In some embodiments, a field programmable gate array may interact with a microprocessor to perform any of the methods described herein. In general, in some embodiments, the methods are performed by any hardware device. This may be general-purpose hardware such as a computer processor (CPU) or hardware specific to the method such as an ASIC.

The above-described embodiments are merely illustrative of the principles of the present invention. It is understood that modifications and variations of the arrangements and details described herein will occur to others skilled in the art. Therefore, it is intended that the invention be limited only by the scope of the following claims and not by the specific details presented in the description and explanation of the embodiments herein.

bibliography Scientific references

• M. Sahani, A. Nayak, R. Agrawal and D. Sahu, "A GSM, WSN and embedded web server architecture for Internet based kitchen monitoring system," 2015 International Conference on Circuits, Power and Computing Technologies [ICCPCT-2015], Nagercoil, India, 2015, pp. 1-6, doi: 10.1109/ICCPCT.2015.7159480 .
• H. Pudugosula, "Automatic Smart and Safety Monitoring System for Kitchen Using Internet of Things," 2019 International Conference on Intelligent Computing and Control Systems (ICCS), Madurai, India, 2019, pp. 1174-1177, doi: 10.1109/I CCS45141 .2019.9065663 .
• Žarić, Nikola, Milutin Radonjić, Nikola Pavlićević, and Sanja Paunović Žarić. 2021. "Design of a Kitchen-Monitoring and Decision-Making System to Support AAL Applications" Sensors 21, no. 13: 4449. https://doi.org/10.3390/s21134449
• Hassan, Ch Anwar Ul, Jawaid Iqbal, Muhammad Sufyan Khan, Saddam Hussain, Adnan Akhunzada, Mudabbir Ali, Abdullah Gani, Mueen Uddin, and Syed Sajid Ullah. 2022. "Design and Implementation of Real-Time Kitchen Monitoring and Automation System Based on Internet of Things" Energies 15, no. 18: 6778. https://doi.org/10.3390/en15186778
• Umapathi, N., Sabbani, S. (2022). An Internet of Things (loT)-based Approach for RealTime Kitchen Monitoring Using NodeMCU 1.0. In: Sivasubramanian, A., Shastry, P.N., Hong, P.C. (eds) Futuristic Communication and Network Technologies. VICFCNT 2020. Lecture Notes in Electrical Engineering, vol 792. Springer, Singapore. https://doi.org/10.1007/978-981-16-4625-6 4
• Hsu, Wei-Ling, Ji-Yun Jhuang, Chien-Shiun Huang, Chiu-Kuo Liang, and Yan-Chyuan Shiau. 2019. "Application of Internet of Things in a Kitchen Fire Prevention System" Applied Sciences 9, no. 17: 3520. https://doi.org/10.3390/app9173520
• F Nugroho and A B Pantjawati, "Automation and Monitoring Smart Kitchen Based on Internet of Things (loT)", IOP Conference Series: Materials Science and Engineering, Volume 384, International Symposium on Materials and Electrical Engineering (ISMEE) 2017, Bandung, Indonesia

Patents

• CN212645834U ; Overflow monitoring device
• CN216221393U ; Water overflow detection device
• CN203799469U ; Overflow alarm device
• US20130179159A1 ; System and methods for detecting overflow
• CN101957083A ; Multifunctional overflow device
• CN217090649U ; Dishwasher
• CN112804600A ; Sound equipment based on AI artificial intelligence and system thereof
• US20190354815A1 ; Homeostatic Capacity Evaluation of Artificial Intelligence Systems
• US20210232670A1 ; Artificial Intelligence service method and device therefor
• CN112362750A ; Acoustic detection system and device based on artificial intelligence
• CN214011777U ; Artificial intelligence test system
• CN207731301U ; A kind of intelligent sensor system framework
• CN105659629A ; Ear-shaped part, artificial head part, and measurement system and measurement method using these
• CN111496802A ; Control method, device, equipment and medium for artificial intelligence equipment
• CN111161727A ; Artificial intelligence barrier-free control system
• GB2588036A ; Sound monitoring system and method
• GB2563892A ; Sound monitoring system and method
• US20060177071A1 ; Method and system for detecting a predetermined sound event such as the sound of breaking glass
• CN103197630B ; The self study criterion method of formation of Audio Monitoring System
• KR20180066294A ; Sound monitoring system and method using multiple devices
• JP3693644B2 ; Equipment operating state acoustic monitoring method and equipment operating state acoustic monitoring apparatus
• US8155326B2 ; System, method, and apparatus for using the sound signature of a device to determine its operability
• EP1913567A2 ; Enhanced acoustic monitoring and alarm response
• JP2013225248A ; Sound identification system, sound identification device, sound identification method and program.

literature

• [1] " How do Germans cook?", GfK, 2015. Accessed: November 21, 2023. [Online]. Available at: https://aus-meinem-kochtopf.de/wie-kochen-die-deutschen/
• [2] C. Wan, "How We Can Enable Intelligence On Endpoints," ambiq. Accessed: December 12, 2023. [Online]. Available at: https://ambiq.com/blog/how-we-canenable-intelligence-on-endpoints/
• [3] " The 5 most common causes of fire", Fire Protection Central, May 2022. Accessed: December 15, 2023. [Online]. Available at: https://brandschutz-zentrale.de/brandschutz/einsatz/die-5-haeufigsten-brandursachen/
• [4] H.-H. Drews, "Every fifth fire in the home starts on the stove! Cause: people", 2017. Accessed: July 10, 2023. [Online]. Available at: https://www.schadenprisma.de/wp-content/uploads/sp_2017_4_2.pdf
• [5] " MP3 developers enter the hall of fame", stern, March 2007. Accessed: November 15, 2023. [Online]. Available at: https://www.stern.de/digital/computer/-cehall-of-fame--mp3-entwickler-kommen-in-ruhmeshalle-3356668.html
• [6] Patent applications: TU Ilmenau among the top ten nationwide. 2023. Accessed: October 18, 2023. [Radio report]. Available at: https://www.mdr.de/nachrichten/thueringen/sued-thueringen/ilmenau-ilmkreis/tu-ilmenau-innovativ-patentiwstudie-100.html
• [7] " Dipl.-Ing. Sascha Grollmisch". Accessed: November 22, 2023. [Online]. Available at: https://www.idmt.fraunhofer.de/en/institute/doctorands/grollmisch.html#2
• [8th] " Cooking methods - boiling, steaming, stewing &Co." Accessed: November 13, 2023. [Online]. Available at: https://www.verbraucherservice-bayern.de/themen/hauswirtschaft/garmethoden-kochen-daempfen-duensten-co#:-:text=Ob%20Sieden%2C%20Dünsten%2C%20Braten%20oder,eine%20enorme%20Vielfalt%20beim%20Kochen
• [9] Wikipedia authors, " Noise", Wikipedia. May 30, 2023. Accessed: October 30, 2023. [Online]. Available at: https://de.wikipedia.org/wiki/Geräusch
• [10] Wikipedia authors, " Spectrogram", Wikipedia. July 28, 2023. [Online]. Available at: https://de.wikipedia.org/wiki/Spektrogramm
• [11] Wikipedia authors, " Short-time Fourier transform", Wikipedia. November 21, 2023. Accessed: December 15, 2023. [Online]. Available at: https://de.wikipedia.org/wiki/Kurzzeit-Fourier-Transformation
• [12] S. Jauss, "Fundamentals of signal processing and audio features for machine learning", AI in the audio sector. Accessed: December 15, 2023. [Online]. Available at: https://ai.hdm-stuttgart.de/news/2023/ki-im-audiobereich-grundlagen-signalverarbeitung-ml/
• [13] L. Monigatti, "Audio Classification with Deep Learning in Python," Towards Data Science. Accessed: 10 May 2023. [Online]. Available at: https://towardsdatascience.com/audio-classification-with-deep-learning-in-python-cf752b22ba07
• [14] K. Nighania, "Various ways to evaluate a machine learning model's performance," Towards Data Science. Accessed: 15 December 2023. [Online]. Available at: https://towardsdatascience.com/various-ways-to-evaluate-a-machine-learning-models-performance-230449055f15
• [15] " AI-based acoustic monitoring in production". Accessed: December 15, 2023. [Online]. Available at: https://www.idmt.fraunhofer.de/de/business_units/industrial-sound-analysis.html
• [16] C. Nugteren, "tinyML Talks: Demoing the world's fastest inference engine for Arm Cortex-M (from second 822)", January 6, 2022. Accessed: November 22, 2023. [Online]. Available at: https://www.youtube.com/watch?v=ComEgcN7KfY&t=822s
• [17] " tinyML Summit 2021 Proceedings", 22 March 2021. Accessed: 31 October 2023. [Online]. Available at: https://cms.tinyml.org/wp-content/uploads/summit2021/tinyMLSummit2021d1 Awards_Syntiant.pdf
• [18] " High-quality audio signal acquisition", Trend Guide Sensor Technology. Accessed: December 15, 2023. [Online]. Available at: https://www.infineon.com/dgdl/lnfineon-Xensiv%20MEMS%20Mikrofone_MarktundTechnik-ART-v01_00-DE.pdf?fileId=5546d46264fee02f01650e895f1b4854#:~:text=MEMS%2DMicrofones%20offer%20advantages%20over%20her,multiple%20microphones%20matching%20to%20each%20other.
• [19] " Smart Kitchen Appliances Market Size, Share & Trends Analysis Report By Product (Smart Refrigerators, Smart Cookware & Cooktops, Smart Dishwashers, Smart Oven, Others), By Application, By Region, And Segment Forecasts, 2023 2030", Grand View Research, Report Overview, 2022. Accessed: September 20, 2023. [Online]. Available at: https://www.grandviewresearch.com/industry- ana!y-sis/smart-kitchen-appliances-market
• [20] " Smart Kitchen Appliance Market Forecast to 2028 COVID-19 Impact and Global Analysis By Product (Ovens, Refrigerators, Sous Vide, Juicers & Blenders, Cooker & Cooking Robots, Cooktops, & Integrated Ovens & Cooktops, Others), by End User, and Connectivity", Research And Markets, Report Overview, 2021. [Online]. Available at: https://www.researchandmarkets.com/reports/4787 380/smart-kitchen-appliance-market-forecast-to?utm_source=GNOM&utm_medium=PressRelease&utm_code=qr738s&utm_cam paign=1693860+-+Smart+Kitchen+Appliance+Global+Market+Forecast%3a+An+E-stimated+US%24+47%2c071+Million+Market+by+2028%2c+with+CAGR+of+17.4%2 5+Forecast+During+20202028&utm_exec=cari 18prd
• [21] " Induction Cooktops Market Size, Share & Trends Analysis Report By Product (Builtin, Free-standing), By Application (Household, Commercial), By Distribution Channel (Online, Specialty Stores), And Segment Forecasts, 2021 2028", Grand View Research, Report Overview, 2019. [Online]. Available at: https://www.grandviewresearch.com/industry-analysis/induction-cooktops-market
• [22] Statista, "Cooking appliances and equipment (INDUSTRIES &MARKTS)", Statista, Statistic report did-23654-1, Nov. 2022. Accessed: 16 September 2023. [Online]. Available at: https://www.statista.com/study/23654/cooking-appliances-andequipment-statista-dossier/
• [23] " Almost 6 million older people live alone", Federal Statistical Office (Destatis), Sep. 2021. [Online]. Available at: https://www.destatis.de/DE/Presse/Pressemitteilungen/2021/09/PD21_N057 _12411.html
• [24] " GERMANY KITCHEN APPLIANCES MARKET SIZE & SHARE ANALYSIS GROWTH TRENDS & FORECASTS (2023 2028)", Mordor Intelligence, Report Overview, 2022. Accessed: 25 September 2023. [Online]. Available at: https://www.mordorintelligence.com/industry-reports/germany-kitchen-appliancesproducts-market-industry
• [25] A. Viroli, L. Jeanneteau, M. Nostro, and SE Christiansen, "Induction hob with boiling detection and induction energy control, method for heating food with an induction hob and computer program product ", EP3195694A1 Accessed: 26 July 2023 . [Online]. Available at: https://patents.google.com/patent/EP3195694A1/en
• [26] C. Franco Gutierrez et al., "Cooking appliance device ", EP2590473A1 Accessed: 26 July 2023 . [Online]. Available at: https://patents.google.com/patent/EP2590473A1/de?oq=EP2590473A1
• [27] AR Buendia Garcia, F. Ashrafzadeh, and M. MUKHTAR, "Boil and boil-dry detection methods for cooking appliances using vibration sensors ", US9354207B2 Accessed: July 24, 2023 . [Online]. Available at: https://patents.google.com/patent/US9354207B2/en?oq=US9354207B2
• [28] G. Dr. med. Betz, H. Dr. med. Renz-Polster, and A. Dr. med. Schäffler, "Sensory organs and body perception in old age", Gesundh. Heute, 2014, accessed: October 18, 2023. [Online]. Available at: https://www.apotheken.de/krankheiten/hintergrundwissen/10299-sinnesorgane-und-koerperwahrnehmung-im-alter
• [29] S. Ahrens, "Per capita consumption of pasta in Germany until 2021/22", Statista, Sep. 2023. Accessed: September 26, 2023. [Online]. Available at: https://de.statista.com/statistik/daten/studie/156453/umfrage/pro-kopf-verbrauchvon-nudeln-in-deutschland-seit-1996/
• [30] " Wind energy How much electricity does a wind turbine produce?", Bluedrift Knowledge and Information. Accessed: September 12, 2023. [Online]. Available at: https://www.bluedrift.at/index.php/wissen-und-infos/windkraft
• [31] L. Fernändez, "Weighted average cost for installed onshore wind energy worldwide from 2010 to 2021", Statista, Statistics, Feb. 2023. Accessed: 26 September 2023. [Online]. Available at: https://www.statista.com/statistics/506774/weightedaverage-installed-cost-for-onshore-wind-power-worldwide/
• [32] R. Bocksch, "Let there be light (ELECTRICITY)", May 2023. Accessed: September 18, 2023. [Online]. Available at: https://de.statista.com/infografik/21802/anteilder-bevoelkerung-mit-zugang-zu-elektrizitaet/
• [33] " Everything you need to know about hobs", Küchen Staude online magazine. Accessed: September 19, 2023. [Online]. Available at: https://www.kuechen-staude.de/magazin/highlights/kochfelder-az/
• [34] " Germany households and household members (2022)", Federal Statistical Office (Destatis), March 2023. [Online]. Available at: https://www.destatis.de/DE/Themen/Gesellschaft-Umwelt/Bevoelkerung/Haushalte-Familien/Tabellen/1-1-privathaushalte-haushaltsmitglieder.html
• [35] Country data, " Energy budget in Germany", country data, 2021. Accessed: 19 September 2023. [Online]. Available at: https://www.laenderdaten.info/Europa/Deutschland/energiehaushalt.php
• [36] " Europe Population 2022", Federal Statistical Office (Destatis), March 2023. [Online]. Available at: https://www.destatis.de/Europa/DE/Thema/Basistabelle/Bevoelkerung.html
• [37] B. Urmersbach, "European Union & Eurozone: Number of private households from 2012 to 2022", Statista, July 2023. Accessed: September 19, 2023. [Online]. Available at: https://de.statista.com/statistik/daten/studie/349055/umfrage/privathaushalte-in-eu-und-euro-zone/#:~:text=lm%20Jahr%202021%20gab%20es,es%20rund%20153%2C4%20 Millionen
• [38] A.-S. Turulski, "World population by continent in mid-2022", Statista, May 2023. Accessed: September 19, 2023. [Online]. Available at: https://de.statista.com/statistik/daten/studie/1723/umfrage/weltbevoelkerung-nach-kontinenten/
• [39] United Nations, "Household size and composition around the world", United Nations Department of Economic and Social Affairs, 2017. Accessed: September 16, 2023. [Online]. Available at: https://www.un.org/en/development/desa/population/publications/pdf/popfacts/PopFacts_2017-2.pdf
• [40] R. Muschter, "Average household size in China from 1990 to 2021", Statista, March 2023. Accessed: September 19, 2023. [Online]. Available at: https://de.statista.com/statistik/daten/studie/220587/umfrage/durchschnittliche-groesse-der-haushalte-in-china/#:~:text=Durchschnittliche%20Größe%20der%20Haushalte%20in%20China%20bis%202021&text=lm%20Jahr%202021%20hat%20die,d en%20Jahren%201990%20bis%202021

Claims (17)

Cooking appliance (10), having the following features: a heat source (12) for heating a medium or a cooking vessel; Sound detection means (54) designed to receive an acoustic signal originating from the medium or the cooking utensil and/or to convert it into a digital acoustic signal; a processor (55) configured to analyze the digital acoustic signal with respect to raw data or one or more features and to determine a cooking state for the medium based on the one or more features. Cooking appliance (10) according to claim 1, wherein the processor (55) is designed to recognize one or more cooking states, in particular predefined cooking states, based on one or more patterns of features and/or combinations of features. Cooking appliance (10) according to one of the preceding claims, wherein the cooking state is selected from the group of cooking states comprising one of the following: - Temperature information of the medium; - Boiling information of a fluid as a medium; - evaporation/combustion state of a fat as a medium; and - Cooking status of the medium and/or of a food item as a medium. Cooking appliance (10) according to one of the preceding claims, wherein the cooking appliance (10) has the form of a hob; and/or wherein the cookware comprises a pot or a pan and/or wherein the medium comprises a food (e.g. meat) and/or a cooking medium (e.g. cooking water or fat); and/or wherein the cooking appliance (10) has the form of a hob comprising a plurality of cooking plates or cooking positions; wherein the processor (55) is designed to carry out a cooking position detection or a cooking plate detection based on the digital acoustic signal or the digital acoustic signals and/or additional information, wherein the cooking position or the cooking plate detection information represents a feature. Cooking appliance (10) according to one of the preceding claims, wherein the means for sound detection (54) are designed to be aligned or to align themselves with a respective cooking utensil or medium associated with a cooking position or hotplate. Cooking appliance (10) according to one of the preceding claims, wherein the means for sound detection (54) comprise means for sound detection (54) of airborne sound, a microphone or microphone array and/or means for sound detection (54) of structure-borne sound; and/or
wherein the sound detection means (54) comprise external sound detection means (54) which are arranged on the cooking appliance (10) and/or on the cookware and/or the kitchen ceiling in the region of the hob and/or the extractor hood. Cooking appliance (10) according to one of the preceding claims, wherein in analyzing the processor (55) performs one or more of the following operations: • A/D conversion of the acoustic signals • Preprocessing (e.g. including steps such as quantization or normalization) • Feature extraction, e.g. in the form of a time-frequency transformation, or extraction of other predefined features; • Signal filtering, in particular for minimizing noise or minimizing noise originating from the heating source (induction) (12), for example in the form of a bandpass filter; • Beamforming Cooking appliance (10) according to one of the preceding claims, wherein the processor (55) is designed to carry out the analysis by means of machine learning and/or a neural network. Cooking appliance (10) according to claim 8, wherein the processor (55) is designed to train the neural network and/or the ML algorithm by means of training data and/or by means of a current digital acoustic signal. Cooking appliance (10) according to claim 9, wherein the processor (55) is designed to be trained on site, to be trained in the delivery state, to be adapted on site and/or to be trained by means of federated training. Cooking appliance (10) according to one of the preceding claims, wherein the processor (55) is designed to post-process a signal derived from the digital acoustic signal by means of privacy enhancement processing and/or by means of a speech filter. Cooking appliance (10) according to one of the preceding claims, wherein the processor (55) is designed to carry out the analysis based on further external features, in particular an excitation frequency of the cooking utensil, a temperature of the cooking utensil, pot position information, a pot detection state, a cooking appliance setting, and/or information about the medium (food to be cooked and/or cooking medium). Cooking appliance (10) according to one of the preceding claims, which has an internal database and/or an interface to an external database, wherein model weightings and/or information relating to one or more reference features associated with one or more cooking states are stored on the internal database or the external database and wherein the processor (55) is designed to carry out an analysis using the one or more reference features. Cooking appliance (10) according to one of the preceding claims, wherein the cooking appliance (10) has a control which is designed to regulate the heating source (12) as a function of a determined cooking state, in particular a power for the heating source (12) when a cooking state, in particular a boiling state, is reached. Cooking accessory for use with a cooking appliance (10) having a heat source (12) for heating a medium or a cooking vessel, having the following features: Sound detection means (54) designed to receive an acoustic signal originating from the medium or the cooking utensil and/or to convert it into a digital acoustic signal; a processor (55) configured to analyze the digital acoustic signal with respect to raw data or one or more features and to determine a cooking state for the medium based on the one or more features. Method for monitoring a cooking appliance (10) with a heat source (12) for heating a medium or a cooking vessel, comprising the following steps: Receiving an acoustic signal originating from the medium or the cookware by means of sound detection means and/or converting it into a digital acoustic signal; Analyzing the digital signal with respect to raw data or the one or more characteristics and determining a cooking condition based on the one or more characteristics. Computer program with a program code for carrying out the method according to claim 16, when the program code is executed on a processor (55).

Images