49 results about "Audio watermark" patented technology

The invention discloses an audio watermark embedding method, an audio watermark extracting method and a model training method, and belongs to the technical field of artificial intelligence. The audio watermark embedding method comprises the following steps: acquiring watermark data and an original audio clip in which a watermark needs to be embedded in original audio data; inputting the original audio clip and the watermark data into an encoder, and determining embedding coefficients of at least two time frames in the original audio clip through the encoder; based on the embedding coefficient, the watermark data is embedded into the original audio clip, a watermark-containing audio clip corresponding to the original audio clip is obtained, and the embedding coefficient of the time frame is in positive correlation with the energy of the time frame; and replacing the original audio clips in the original audio data with the audio clips containing the watermarks to obtain the audio data containing the watermarks.

The invention discloses a training method and device of a reversible neural network for audio watermark processing. The method comprises the steps of obtaining a sample data set; repeatedly executing the following process until the target loss value is smaller than a preset threshold value, stopping iteration, and obtaining a target reversible neural network: adding watermark information to each piece of sample audio data in the target training set according to an encoder of the initial reversible neural network, and obtaining a plurality of pieces of first sample audio data embedded with the watermark information; performing simulation attack on the plurality of first sample audio data according to a weighted attack strategy to obtain a plurality of second sample audio data; decoding the plurality of second sample audio data according to a decoder of the initial reversible neural network to obtain restored watermark information of each second sample audio data; and determining a target loss value according to the loss values corresponding to all the sample audio data, and adjusting network parameters of the initial reversible neural network according to the target loss value.

Described herein is a system for performing watermark detection using multiple time shifts to increase a resolution of watermark detection. Instead of decoding blocks of successive audio frames (e.g., 10 ms of audio) using a single watermark decoder, watermark verification can be performed by decoding overlapping frame shifts using multiple decoders in parallel, thereby increasing a chance that one of the watermark decoders will be synchronized with the embedded audio watermark. For example, watermark verification may split audio data into parallel streams and decode using two decoders (e.g., 2× shifts-per-frame), four decoders (e.g., 4× shifts-per-frame), or the like. Increasing resolution by performing overlapping detection increases an accuracy of the watermark detection without changing the embedded audio watermark.

A method includes adding, by a first computing device, a first audio watermark to first speech data corresponding to playback of a first utterance including a hotword used to invoke an attention of a second computing device. The method includes outputting, by the first computing device, the playback of the first utterance corresponding to the watermarked first speech data. The second computing device is configured to receive the watermarked first speech data and determine to cease processing of the watermarked first speech data.

Methods, systems, and user devices for suppressing hotword triggering when a hotword in a recorded media is detected are disclosed, including computer programs encoded on computer storage media. In an aspect, a method includes receiving, at data processing hardware, audio data corresponding to playback of a media content item, the audio data including an audio watermark and an utterance of a command preceded by a hotword; determining, by the data processing hardware, that the received audio data includes the hotword; processing, by the data processing hardware, the audio data to: identify the audio watermark included in the audio data; and determine a corresponding bitstream of the audio watermark; and based on the determined corresponding bitstream of the audio watermark, determining, by the data processing hardware, to bypass performing the command preceded by the hotword without accessing an audio watermark database to identify a matching audio watermark.

The invention provides an audio watermark embedding method and device and a conference system, and relates to the technical field of digital watermarking. And obtaining an original audio collected from an environment where the terminal is located in the first time period, wherein the original audio comprises an audio played by the terminal in the first time period. The playing audio is configured to be embedded with a watermark based on the first watermark embedding parameter value by using a first watermark embedding algorithm. And carrying out watermark extraction on the original audio according to the first watermark embedding algorithm. And if the watermark extraction from the original audio fails, embedding the watermark in the audio to be played of the terminal based on the second watermark embedding parameter value by using the first watermark embedding algorithm to obtain a new played audio. Under the condition that the watermark cannot be successfully extracted from the original audio, the watermark embedding parameter value is adjusted, so that the watermark embedded in the audio played later by the terminal can adapt to the environment and is successfully extracted, the adaptive environment adjustment of the audio watermark is realized, and the robustness of the audio watermark is improved.

Described is a neural network based system for use in audio watermarking, comprising: a neural network based watermark generator configured for embedding, based on a learnable embedding table, a predefined watermark into an input audio signal, thereby obtaining a watermarked audio signal; and a neural network based watermark detector comprising: a detection head that is configured for determining presence or absence of a watermark in a target audio signal; and a decoding head that is configured for decoding the watermark from the target audio signal, wherein the embedding table is shared between the watermark generator and the watermark detector, such that the decoding of the watermark from the target audio signal is based on the embedding table that has been used for the embedding.

The invention discloses an audio watermark generation and detection method. The method comprises the following steps: acquiring an audio signal corresponding to an audio to be processed; the amplitude of the confrontation disturbance signal is smaller than that of a value set corresponding to a psychological acoustic masking threshold value of the audio signal, the psychological acoustic masking threshold value is used for quantitatively representing an added disturbance amplitude upper limit corresponding to each frequency point in the audio signal, and the confrontation disturbance signal is used for carrying watermark information of the audio to be processed; and embedding the adversarial disturbance signal into the audio signal. The technical problem that the audio watermarking technology adopted in the related technology is mostly based on a static or dominant embedding mode and is easily removed by an attacker through a signal processing method is solved.

The invention discloses a robust audio watermarking method based on an adaptive quantization strategy and feature classification, and belongs to the technical field of digital audio copyright protection. The method comprises the following steps: framing an audio signal, extracting a logarithmic mean feature (DWT-CLM) of discrete wavelet transform, and combining a zero-crossing rate, a variance and energy to form a frame feature vector; a Sigmoid classifier is used to discriminate frame characteristics, and a fixed or variable quantization step size is adaptively selected to embed watermark bits into approximate components; and during extraction, the watermark is accurately extracted through the same feature analysis and classifier discrimination recovery quantization mode. Experiments show that the algorithm has better inaudible property and robustness, can effectively resist attacks such as MP3 compression, resampling, low-pass filtering and re-recording, and is suitable for digital audio copyright protection scenes.

The invention relates to an audio watermark embedding and extracting method for resisting a desynchronization attack, which comprises the following steps of: bearing an embedded bit by using three sections of intermediate-frequency average energy at an embedded end, implementing a triangular modulation strategy on the three sections of average energy according to the embedded bit, and introducing buffer compensation at the embedded end; uniformly scaling all frequency domain coefficients according to a proportion for the buffer expansion interval by taking a segment as a unit, carrying out inverse transformation on the whole frame to obtain frames containing watermarks, and splicing the frames containing the watermarks into audio containing the watermarks; judgment is only carried out on a target frequency band at an extraction side, clipping detection is carried out through symmetric consistency discrimination, rapid and reliable resynchronization in a clipping scene is realized in combination with double-end sliding window search, and blind detection is realized on the premise of not depending on original audio. Compared with the prior art, the method has the advantages that the balance among the watermark capacity, the robustness and the hearing feeling during desynchronization attack resistance is realized, the robustness is improved, and the like.

Embodiments of the present application disclose an audio watermark adding method, comprising: a playing terminal acquiring a first audio in real time; the playing terminal embedding an audio watermark in the first audio, the audio watermark being associated with the playing terminal; and the playing terminal playing the first audio with the embedded audio watermark. Embodiments of the present application also provide an audio watermark analysis method and device and a medium, in a scenario of playing an audio in real time, a playing terminal adds an audio watermark in an audio stream in real time, so that a later device can determine the playing terminal according to the audio watermark when analyzing the watermark, and tracing is facilitated after the first audio is transcribed.

This invention provides a method for processing audio watermarks and an apparatus for generating audio watermarks. The insertion position of a reference code in an initial watermark sequence is determined based on the signal power of the main audio signal to generate an extended watermark sequence. The main audio signal and the extended watermark sequence are then synthesized to generate an audio signal with an embedded watermark. This overcomes noise interference.

The application discloses a method and device for training a reversible neural network for audio watermark processing. The method comprises: obtaining a sample data set; repeatedly performing the following process until the target loss value is less than the preset threshold, stopping iteration, and obtaining a target reversible neural network: adding watermark information to each sample audio data in the target training set according to the encoder of the initial reversible neural network to obtain a plurality of first sample audio data embedded with watermark information; performing simulated attacks on the plurality of first sample audio data according to a weighted attack strategy to obtain a plurality of second sample audio data; decoding the plurality of second sample audio data according to the decoder of the initial reversible neural network to obtain restored watermark information of each second sample audio data; determining the target loss value according to the loss value corresponding to all sample audio data, and adjusting the network parameters of the initial reversible neural network according to the target loss value.

A method and system for adding overtly human-perceptible supplemental audio content into a media stream to help mask audio effects of an audio watermark in the media stream. A method involves receiving a media stream that defines a sequence of audio content presentable by a content presentation device, modifying the media stream to produce a modified media stream that defines the sequence of audio content, and outputting the modified media stream for presentation by the content presentation device. The modified media stream includes an audio watermark that is machine-detectable to trigger an interactive event. Further, the act of modifying the media stream involves adding into the media stream supplemental audio content coincident with the audio watermark, to help mask the audio watermark in the modified media stream during presentation of the modified media stream by the content presentation device.

Disclosed example people monitoring methods include detecting a first watermark in a first audio signal obtained from an acoustic sensor, the first watermark identifying media presented by a monitored media device, determining whether a second watermark, different from the first watermark, is embedded in the first audio signal obtained from the acoustic sensor, the second watermark identifying at least one of a mobile device or a user of the mobile device, classifying the second watermark as a media watermark or a people monitoring watermark based on a characteristic of the second watermark, and when the second watermark is determined to be embedded in the first audio signal, reporting at least one of the second watermark or information decoded from the second watermark to identify at least one of the mobile device or the user of the mobile device as being exposed to the media presented by the monitored media device.

The application provides a multi-bit audio watermarking method based on phase distribution and high-efficiency bit mapping, and belongs to the field of audio information hiding. The application divides multiple independent phase subintervals based on the natural distribution characteristics of phases, and can realize one-to-many mapping from single feature to multi-bit watermarking by constructing unique phase features in the phase subintervals corresponding to watermarking information. In addition, the application introduces an optimization strategy, and can realize automatic phase redistribution. The application can significantly reduce the number of subspaces / feature modes required for multi-bit embedding, and reduce the design difficulty of the multi-bit watermarking algorithm. In addition, since an automatic phase redistribution algorithm based on optimization constraints is adopted, the algorithm can more effectively balance the inaudibility and robustness.

The invention provides an audio watermark embedding method and device, electronic equipment and a storage medium, and belongs to the technical field of audio processing, and the method comprises the steps: obtaining the amplitude spectrum characteristics of an original audio signal, and determining a watermark embedding energy mask; inputting the amplitude spectrum features into a plurality of mask prediction network models to obtain watermark embedding risk masks; determining a target watermark embedding risk mask according to the watermark embedding risk mask to generate a frame-level embedding mask; and embedding the watermark information into the feature component corresponding to the target audio frame to obtain a watermark-containing amplitude spectrum feature, and performing inverse time-frequency transformation on the watermark-containing amplitude spectrum feature to obtain a watermark-containing audio signal. According to the method, the audio energy and the model prediction risk are combined, the audio frame which is most suitable for embedding the watermark is adaptively screened out, modification in hearing sensitive frames such as silent frames and transient frames is avoided, the problem that the audio quality is reduced due to undifferentiated embedding of the watermark is solved, and on the premise that the watermark robustness is not sacrificed, the watermark embedding efficiency is improved. And the imperceptibility of the watermark and the auditory quality of the audio containing the watermark are obviously improved.

The invention discloses a high-robustness lossless high-capacity audio watermark embedding method based on deep learning, and relates to the technical field of digital watermarking. The method comprises the following steps of audio-watermark information preprocessing, watermark information embedding and watermark embedding detection and judgment. According to the method, the original audio and the watermark information to be embedded are correspondingly preprocessed, so that the quality of the original audio and the watermark information is effectively improved, the subsequent embedding process is more efficient and accurate, then the preprocessed audio and watermark information are input into the audio embedding network to output the watermark audio, and the watermark embedding efficiency is improved. According to the method, high-capacity watermark information is embedded, audio distortion in the embedding process is effectively avoided, the quality of the embedded audio is guaranteed, finally, the stability and the anti-jamming capability of the embedded watermark can be effectively detected by introducing adversarial training attack simulation, the audio distortion is reduced to the maximum extent while the watermark capacity is guaranteed, and the watermark quality is improved. And the anti-interference capability of the watermark is improved.

A method includes adding, by a first computing device, a first audio watermark to first speech data corresponding to playback of a first utterance including a hotword used to invoke an attention of a second computing device. The method includes outputting, by the first computing device, the playback of the first utterance corresponding to the watermarked first speech data. The second computing device is configured to receive the watermarked first speech data and determine to cease processing of the watermarked first speech data.

Embodiments of the present application provide an audio watermark embedding method and device, electronic equipment and storage medium, belonging to the technical field of speech processing. The method comprises: performing attribute division on watermark attribute data to obtain a watermark attribute block, and performing content division on watermark content data to obtain a watermark content block; performing encoding processing on the watermark attribute block, merging the encoded attribute code sequence block and the watermark content block to obtain a target watermark data block; and embedding the target watermark data block into a target spectrum according to an encoding rate to obtain a target audio. Through encoding the watermark attribute block, the embodiments of the present application can enable the watermark to be identified even when the position of the watermark in the audio signal is shifted; and embedding the target watermark data block into the target spectrum according to the encoding rate can repeatedly embed the audio watermark into the audio, and when part of the audio watermark is damaged or lost, complete watermark information can still be extracted from the remaining audio, thereby improving the effect of audio watermark embedding.

An audio watermark embedding method (600) and apparatus, and a conference system, relating to the technical field of digital watermarking. The method (600) comprises: acquiring first original audio collected, during a first time period, from an environment where a first terminal is located, the first original audio comprising first playback audio played back by the first terminal during the first time period, and the first playback audio being configured to use a first watermark embedding algorithm to embed a first watermark on the basis of a first watermark embedding parameter value (601); performing watermark extraction on the first original audio on the basis of the first watermark embedding algorithm (602); and if the extraction of the first watermark from the first original audio fails, using the first watermark embedding algorithm to embed, on the basis of a second watermark embedding parameter value, the first watermark into first audio to be played back that is to be played back by the first terminal, so as to obtain second playback audio, the second watermark embedding parameter value differing from the first watermark embedding parameter value, and the second playback audio being configured to be played back by the first terminal during a second time period (603).

Techniques are described for hybrid masking threshold-based perceptual slacks for audio watermarking. In some embodiments, the techniques include identifying an audio signal, determining perceptual slacks for the audio signal, generating a watermarked audio signal that includes an audio watermark based on the perceptual slacks, and outputting the watermarked audio signal using one or more speakers, for localization of the one or more speakers.

The invention discloses an identity verification method based on an audio watermark resistant to digital-to-analog conversion and related equipment, and relates to the technical field of financial science and technology and information security. According to the scheme, watermark extraction is carried out in an audio to be subjected to identity verification in which a scrambled and encrypted client identity verification information watermark is embedded through a digital audio watermark embedding algorithm; as the watermark information is scrambled and encrypted, the security and customer privacy of the watermark information in the transmission and extraction process are ensured. A digital audio watermark resistant to digital-to-analog conversion is applied to an identity verification scene when a customer handles business by using remote banking service, and the identity verification of the customer is realized by embedding and extracting the digital audio watermark in the scene, so that the security of the remote banking service and the customer privacy protection level are improved. Due to the digital audio watermarking based on digital-to-analog conversion resistance, the accuracy of extracting and identifying the watermarking can be ensured.

The application relates to the fields of artificial intelligence and financial technology, and discloses an audio watermark embedding method and device, computer equipment and a storage medium, which are characterized in that the method decomposes host audio to be embedded with a watermark into a plurality of subbands of different frequency bands through discrete wavelet transform; an encoder in a U-Net network model is used to perform layer-by-layer down sampling on each of the subbands, and hidden space features of the host audio are extracted; watermark information to be embedded is determined, and the watermark information is encoded to obtain a watermark vector; the watermark vector is embedded in the hidden space features to obtain fusion features; a decoder in the U-Net network model is used to perform layer-by-layer up sampling on the fusion features, and audio information is reconstructed; inverse discrete wavelet transform is performed on the audio information to generate target audio embedded with the watermark information; and therefore, the application can effectively improve the flexibility, adaptability and robustness of audio watermark embedding.

The invention provides a deep robust audio watermarking method based on diffusion model residual block modulation. The method comprises the following steps: firstly, constructing a conditional modulation signal through fusion of a watermark vector and time step embedding; the method is characterized in that a feature-level linear modulation module FiLM is integrated in a residual block of a diffusion noise prediction network, and deep penetration of watermark information in a multi-level feature space is realized through channel-level scaling and offset; in the embedding stage, a deterministic scheduler is utilized to perform forward noise addition on audio, and intensity weighting and superposition are performed on predicted noise obtained by single reasoning in combination with a self-adaptive masking graph generated by local energy distribution. According to the method, the problems of shallow watermark embedding and weak feature coupling are solved, and the robustness under the interference of filtering, compression and the like is remarkably improved.

Embodiments of the present application provide a method for processing sound watermark and a sound watermark generating device. A receiver obtains a call receiving sound signal. A reflection sound signal is generated according to a virtual reflection condition and the call receiving sound signal. The virtual reflection condition includes a positional relationship among the receiver, a sound source and an external object, and the reflection sound signal is a sound signal obtained by simulating a sound emitted by the sound source, reflected by the external object and recorded by the receiver. A phase of the reflection sound signal is offset according to a watermark identifier to generate a watermark sound signal. The watermark sound signal includes the reflection sound signal with the offset phase. Thus, an echo cancellation mechanism at a receiving end can cancel the watermark sound signal through a feedback path, and a voice signal through a call transmission path is not affected.

A system for avoiding double processing using audio watermarking. A decoder inserts an audio watermark during a transient in an audio signal. This avoids the drawbacks of using out-of-band control signals or metadata. The decoder performs: detecting a transient in a first audio signal; transforming a portion related to the transient into a frequency domain to compare a first frequency band of the frequency domain data to a second frequency band of the frequency domain data; when the first frequency band is not related to the second frequency band, the decoder performs processing on the first audio data to generate second audio data; when the first frequency band is related to the second frequency band, the first audio data is used as the second audio data without performing any processing.

The invention discloses an audio watermark generation method and a related device, and relates to the technical field of artificial intelligence, and the method comprises the steps: obtaining an input audio, carrying out the acoustic feature partitioning and hidden space coding of the input audio, obtaining hidden space representation vectors corresponding to a plurality of acoustic feature blocks of the input audio, and carrying out the acoustic feature partitioning and hidden space coding of the input audio; and fusing the hidden space representation vectors respectively corresponding to the plurality of acoustic feature blocks with the watermark information coding vector to obtain watermark embedding vectors respectively corresponding to the plurality of acoustic feature blocks, and decoding and reconstructing the watermark embedding audio containing the watermark information according to the watermark embedding vectors respectively corresponding to the plurality of acoustic feature blocks. According to the method, a large amount of audio redundant information in the acoustic feature block can be removed through hidden space coding, only more critical features capable of better representing the essence of the audio are reserved, the watermark information coding vectors are embedded into the features, the influence of the watermark on the overall characteristics of the audio can be reduced as much as possible, and therefore the watermark is not prone to being perceived, and the hearing feeling is better.

A method and system for adding overtly human-perceptible supplemental audio content into a media stream to help mask audio effects of an audio watermark in the media stream. A method involves receiving a media stream that defines a sequence of audio content presentable by a content presentation device, modifying the media stream to produce a modified media stream that defines the sequence of audio content, and outputting the modified media stream for presentation by the content presentation device. The modified media stream includes an audio watermark that is machine-detectable to trigger an interactive event. Further, the act of modifying the media stream involves adding into the media stream supplemental audio content coincident with the audio watermark, to help mask the audio watermark in the modified media stream during presentation of the modified media stream by the content presentation device.