Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Response to Amendment
2. Claims 1, 15 and 20 are amended. Claim 10 is cancelled. Claims 1-9 and 11-20 are pending.
Response to Arguments
With regards to applicant’s arguments, filed on 12/10/2025, have been fully considered but they are not persuasive. The applicant asserts, with respect to claims 1, 15 and 20, that the combination of Wu and Truong does not teach or suggest: “determining a contrastive loss function based on a first similarity metric between CI data of each CI pair of the plurality of CI pairs in the training dataset," "determining a reconstruction loss function based on a second similarity metric between the original CI data in the training dataset and predicted CI data generated based on the mask in the training dataset," and "determining a total loss function based on an aggregate of the contrastive loss function and the reconstruction loss function,”. Examiner respectfully disagrees.
The combination of Wu and Truong, specifically, Wu teaches a method of a wireless monitoring system. A time series of channel information (CI) of a wireless multipath channel may be obtained. The time series of CI (TSCI) may be extracted from a wireless signal (signal) transmitted between a Type 1 heterogeneous wireless device and a Type 2 heterogeneous wireless device in a venue through the channel. A characteristics and/or a spatial-temporal information (STI, e.g. motion information) of the object and/or of the motion of the object may be monitored based on the TSCI. A task may be performed based on the characteristics and/or STI. The characteristics and/or STI (e.g. motion information) may be monitored individually based on a TSCI associated with a particular Type 1 device and a particular Type 2 device, and/or monitored jointly based on any TSCI associated with the particular Type 1 device and any Type 2 device. A first channel between a Type 1 device and a Type 2 device may be different from a second channel between another Type 1 device and another Type 2 device. (See Wu; Par. [68], [114]-[116]) Wu teaches that a set of selected significant local peaks may be selected from the set of identified significant local peaks based on a selection criterion (e.g. a quality criterion, a signal quality condition). The characteristics/STI of the object may be computed based on the set of selected significant local peaks and frequency values associated with the set of selected significant local peaks. A similarity score and/or component similarity score may be computed (e.g. by a server (e.g. hub device), the processor, the Type 1 device, the Type 2 device, a local server, a cloud server, and/or another device) based on a pair of temporally adjacent CI of a TSCI. [Therefore, the TSCI may be extracted from a motion sensing signal. Type 2 device performs smart sensing tasks, which are based on the trained characteristics [parameters] based on the extracted CI] (See Wu; Par. [164], [194]-[198])
On the other hand, Truong teaches generating a data model for a machine learning application. The data model can be generated using synthetic data in some aspects. This synthetic data can be generated using a synthetic dataset model, which can in turn be generated using actual data. The synthetic data may be similar to the actual data in terms of values, value distributions (e.g., univariate and multivariate statistics of the synthetic data may be similar to that of the actual data), structure and ordering, or the like. Figure 7 describes process 700 for training a classifier for generation of synthetic data. According to process 700, a data sequence 701 can include preceding samples 703, current sample 705, and subsequent samples 707. In some embodiments, data sequence 701 can be a subset of a training sequence. Data sequence 701 may be applied to recurrent neural network 709. Neural network 709 can be configured to estimate whether current sample 705 is part of a sensitive data portion of data sequence 701 based on the values of preceding samples 703, current sample 705, and subsequent samples 707. The similarity metric can depend on a number of elements of the synthetic dataset that match elements of the reference dataset. In some embodiments, the matching can be an exact match, with the value of an element in the synthetic dataset matching the value of an element in the normalized reference dataset. System 100 (In Fig. 1) can compare a synthetic dataset to a normalized reference dataset, a synthetic dataset to an actual (unnormalized) dataset, or to compare two datasets according to a similarity metric. Synthetic data satisfying the similarity criterion can be too similar to the reference dataset. System 100 can be configured to update a loss function for training the generative adversarial network to decrease the similarity between the reference dataset and synthetic datasets generated by the generative adversarial network when the similarity criterion is satisfied. In particular, the loss function of the generative adversarial network can be configured to penalize generation of synthetic data that is too similar to the normalized reference dataset, up to a certain threshold. To that end, a penalty term can be added to the loss function of the generative adversarial network. [Therefore, Truong teaches determining the loss function based on the similarity metric value between synthetic dataset [CI data] and normalized reference dataset [CI pair of the training dataset], determining an updated loss function based on an updated similarity metric value, which is generated by applying the training mask on the original similarity metric, and calculating the total loss function based on the comparison between the original [contrastive] loss function and the updated [reconstruction] loss function] (See Truong; Par. [101], [115], [124]-[126])
Therefore, and for the reasons set above, the combination of Wu and Truong teaches the claimed invention. The rejection of claims 1-9 and 11-20 is sustained.
Claim Rejections - 35 USC § 103
3. The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
This application currently names joint inventors. In considering patentability of the claims the examiner presumes that the subject matter of the various claims was commonly owned as of the effective filing date of the claimed invention(s) absent any evidence to the contrary. Applicant is advised of the obligation under 37 CFR 1.56 to point out the inventor and effective filing dates of each claim that was not commonly owned as of the effective filing date of the later invention in order for the examiner to consider the applicability of 35 U.S.C. 102(b)(2)(C) for any potential 35 U.S.C. 102(a)(2) prior art against the later invention.
4. Claims 1-9 and 11-20 are rejected under 35 U.S.C. 103 as being unpatentable over Wu et al. (US. Pub. No. 2021/0190702 A1) in view of Truong et al. (US. Pub. No. 2020/0012933 A1).
Regarding Claim 1, Wu discloses a method for wireless sensing (See Abstract), comprising:
obtaining channel information (CI) data generated based on at least one wireless channel (See Par. [87], [94], [116], [147] and Fig. 11 of Wu for a reference to type 2 device receives a probe signal in a channel and obtains, at least, one channel information (CI) of the selected channel); generating a training dataset based on the CI data (See Par. [114], [116], [141]-[142], [196] of Wu for a reference to generating a respective series of training probe signals based on the obtained CI or TSCI associated with multiple events), wherein the training dataset comprises: a plurality of CI pairs, original CI data and a mask (See Par. [141]-[144], [179], [194]-[196] of Wu for a reference to an event may be trained based on the projection and the training TSCI associated with the event. The projection maybe retrained using: the training TSCI, at least one current TSCI before re-training the projection and additional training TSCI of the transformation [masked] of the original TSCI);
training a foundation model using the training dataset at least in part by (See Par. [114], [116], [141]-[142], [196] of Wu for a reference to generating a respective series of training probe signals based on the obtained CI or TSCI associated with multiple events),
training a plurality of task-specific models (See Par. [114], [116], [140]-[142] of Wu for a reference to training a classifier of multiple events [Tasks] in a venue based on training TSCI associated with the multiple events. A projection for each CI may be trained based on the training TSCI); and
performing a plurality of wireless sensing tasks based on the foundation model and the plurality of task-specific models (See Par. [88], [145], [166], [295] of Wu for a reference to the TSCI may be extracted from a motion sensing signal. Type 2 device performs smart sensing tasks, which are based on the trained characteristics [parameters] based on the extracted CI), the performing comprises:
generating a feature map using the foundation model based on CI data collected in real-time (See Par. [110], [196], [209]-[210] of Wu for a reference to that based on the obtained CI data over time, a feature map (environmental model) is created), and
inputting the feature map into each of the plurality of task-specific models to perform a corresponding one of the plurality of wireless sensing tasks, respectively (See Par. [110], [196], [209]-[210], [213] of Wu for a reference to the environmental model is inputted the task training models to perform the sensing tasks).
Wu does not explicitly disclose determining a contrastive loss function based on a first similarity metric between CI data of each CI pair of the plurality of CI pairs in the training dataset, determining a reconstruction loss function based on a second similarity metric between the original CI data in the training dataset and predicted CI data generated based on the mask in the training dataset, determining a total loss function based on an aggregate of the contrastive loss function and the reconstruction loss function, and determining model parameters of the foundation model to minimize the total loss function; wherein: each of the plurality of task-specific models is a machine learning model concatenated after the foundation model, the foundation model is a machine learning model having a larger scale than each of the plurality of task-specific models
However, Truong discloses determining a contrastive loss function based on a first similarity metric between CI data of each CI pair of the plurality of CI pairs in the training dataset (See Par. [20], [101], [115], [124] of Truong for a reference to determining the loss function based on the similarity metric value between synthetic dataset [CI data] and normalized reference dataset [CI pair of the training dataset]),
determining a reconstruction loss function based on a second similarity metric between the original CI data in the training dataset and predicted CI data generated based on the mask in the training dataset (See Par. [101], [107], [115], [126] of Truong for a reference to determining an updated loss function based on an updated similarity metric value, which is generated by applying the training mask on the original similarity metric),
determining a total loss function based on an aggregate of the contrastive loss function and the reconstruction loss function (See Par. [101], [115], [124]-[126] and Fig. 7 of Truong for a reference to calculating the total loss function based on the comparison between the original [contrastive] loss function and the updated [reconstruction] loss function), and
determining model parameters of the foundation model to minimize the total loss function (See Par. [61]-[63], [71]-[75] and Fig. 3 of Truong for a reference to determining the training model through model optimizer that optimizes the loss function to the lowest levels); wherein: each of the plurality of task-specific models is a machine learning model concatenated after the foundation model (See Par. [62], [64], [79], [81] and Fig. 1-4 of Truong for a reference to generating a data model for a machine learning application, using synthetic data that can be generated using a synthetic dataset model, which can in turn be generated using actual data. A data model can be trained using computing resources using data provided by dataset), the foundation model is a machine learning model having a larger scale than each of the plurality of task-specific models (See Par. [62], , [81], [112], [126] and Fig. 1-4 of Truong for a reference to the scaling factor of the trained data model *Foundation model) is greater than each data model scaling factor)
Thus, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of Truong to Wu. The motivation for combination would be to improve network’s performance, by improving the generation of machine learning models, through improving the security of sensitive data. (Truong; Par. [7])
Regarding Claim 2, the combination of Wu and Truong, specifically Wu discloses wherein obtaining the CI data comprises:
determining a plurality of device pairs in at least one venue, wherein each of the plurality of device pairs is formed by a transmitter and a receiver (See Par. [64], [152]-[153], [166] of Wu for a reference to determining a pair of wireless device ; Type 1 wireless device, which is a wireless transmitter, and Type 2 wireless device, which is a wireless Receiver);
for each of the plurality of device pairs:
transmitting a wireless signal by the transmitter through a wireless channel (See Par. [48]-[49], [300] and fig. 11 of Wu for a reference to a wireless transmitter is configured to transmit a first wireless signal through a wireless multipath channel in a venue),
receiving the wireless signal by the receiver (See Par. [48]-[49], [300] and fig. 11 of Wu for a reference to a wireless receiver is configured to receive a second wireless signal through the wireless multipath channel in a venue), wherein the received wireless signal differs from the transmitted wireless signal due to the wireless channel and any sensing event in the at least one venue (See Par. [48]-[50] of Wu for a reference to the first transmitted signal differs from the second received signal),
obtaining a time series of channel information (TSCI) of the wireless channel based on the received wireless signal (See Par. [68], [72], [116], [147] and fig. 11 of Wu for a reference to obtaining the time series of channel information [TSCI], which is extracted from the transmitted/received between Type 1 & Type 2 devices in a venue through a channel); and
obtaining the CI data based on all TSCI obtained for the plurality of device pairs (See Par. [116], [147] and fig. 11 of Wu for a reference to extracting the channel info (CI) based on the TSCI obtained from signals exchanged between Type 1 & Type 2 devices).
Regarding Claim 3, the combination of Wu and Truong, specifically Wu discloses wherein generating the training dataset comprises:
processing the CI data to generate preprocessed CI data according to a standardized format readable by the foundation model (See Par. [116]-[118], [142]-[144] of Wu for a reference to processing the channel info (CI) data according to a standardized protocol to generate a preprocessed CI data);
performing a data augmentation on preprocessed CI in the preprocessed CI data to generate augmented CI (See Par. [116]-[118], [142]-[144], [161] of Wu for a reference to generating augmented CI data by processing the preprocessed CI data),
wherein the plurality of CI pairs comprises: a positive CI pair formed by a preprocessed CI and its associated augmented CI, a positive CI pair formed by two preprocessed CI, a positive CI pair formed by two augmented CI, a negative CI pair formed by two CI obtained from two different wireless channels, a negative CI pair formed by two CI obtained from two different venues, a negative CI pair formed by two CI associated with two different sensing events (See Par. [72], [80], [93], [113], [119], [194] of Wu for a reference to the CI data maybe: two different devices using two different channels. The CI pair may be two different values; positive & negative, which are associated with two different sensing tasks).
Regarding Claim 4, the combination of Wu and Truong, specifically Wu discloses wherein processing the CI data comprises:
selecting subcarriers for at least one CI in the CI data to generate a same number of subcarriers for all CI in the CI data according to the standardized format (See Par. [87]-[89] of Wu for a reference to selecting a channel [Subcarrier] from a set of channels. At least, on CI may be obtained by type 2 device from the probe signals transmitted in the selected channel. The same number of channels are selected for all CI data); and
resampling each CI in the CI data to a predetermined temporal rate according to the standardized format (See Par. [119], [122], [177], [200] of Wu for a reference to resampling each CI data to a predetermined spatial-temporal normalized standard format).
Regarding Claim 5, the combination of Wu and Truong, specifically Wu discloses wherein performing the data augmentation comprises:
adding random noise to the preprocessed CI (See Par. [210], [212], [251] and Fig. 9 of Wu for a reference to applying a noise cancellation method by adding noise to the extracted CSI data);
randomizing the selected subcarriers within a block (See Par. [87]-[89] of Wu for a reference to selecting a channel [Subcarrier] from a set of channels. At least, on CI may be obtained by type 2 device from the probe signals transmitted in the selected channel. The same number of channels are selected for all CI data);
performing a time scaling or a time warping on the preprocessed CI (See Par. [181], [305]-[307] of Wu for a reference to applying time scaling of preprocessed CI parameters);
simulating at least one environmental parameter related to multi-path change or occlusion (See Par. [148] of Wu for a reference to simulating multipath channel characteristics [Parameters]); and
normalizing amplitudes of the preprocessed CI to mitigate power variation (See Par. [192], [203], [251] of Wu for a reference to regulating signal amplitude of the processed CI to reduce CIR variation).
Regarding Claim 6, Wu discloses wherein determining the contrastive loss function comprises:
mapping each CI in the training dataset to a corresponding embedding point in an embedding space using the foundation model (See Par. [123], [177]-[178], [199] of Wu for a reference to each CI in the updated/processed data set is mapped to a point in the space);
for each CI pair comprising two CI in the training dataset, generating a distance score between two embedding points corresponding to the two CI of the CI pair based on the first similarity metric (See Par. [132], [155], [194] of Wu for a reference to the similarity score may be based on a pair of adjacent CI from two different TSCI. The similarity score comprises a distance score), wherein:
Wu does not explicitly disclose the distance score is smaller when the CI pair is a positive CI pair, the distance score is larger when the CI pair is a negative CI pair; and
determining the contrastive loss function based on the distance score.
However, Truong discloses the distance score is smaller when the CI pair is a positive CI pair, the distance score is larger when the CI pair is a negative CI pair (See Par. [116]-[118], [122] of Truong for a reference to the distance score is decreased when the CI pair is above zero and is increased when the CI pair is below zero), and
determining the contrastive loss function based on the distance score (See Par. [20], [101], [115], [124] of Truong for a reference to determining the loss function based on the similarity metric value [distance score] between synthetic dataset [CI data] and normalized reference dataset [CI pair of the training dataset]);
Thus, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of Truong to Wu. The motivation for combination would be to improve network’s performance, by improving the generation of machine learning models, through improving the security of sensitive data. (Truong; Par. [7])
Regarding Claim 7, Wu discloses wherein determining the reconstruction loss function comprises:
generating masked CI data at least in part by applying the mask to the original CI data to remove at least one portion of the original CI data along a time dimension or a subcarrier dimension (See Par. [125]-[126], [154] of Wu for a reference to generating the transformed masked CI data by removing a beginning portion and an ending portion of the entire TSCI);
generating the predicted CI data based on the masked CI data using the foundation model (See Par. [188] of Wu for a reference to generating the predicted channel characteristics from the transformed/updated model characteristics);
generating an error function between the original CI data and the predicted CI data based on the second similarity metric (See Par. [196]-[197], [201]-[203] of Wu for a reference to generating an error function between the original channel characteristics and the predicted channel characteristics using similarity metrics).
Wu does not explicitly disclose determining the reconstruction loss function based on the error function.
However, Truong discloses determining the reconstruction loss function based on the error function (See Par. [101], [107], [115], [126] of Truong for a reference to determining an updated loss function based on an updated similarity metric value, which is generated by applying the training mask on the original similarity metric);
Thus, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of Truong to Wu. The motivation for combination would be to improve network’s performance, by improving the generation of machine learning models, through improving the security of sensitive data. (Truong; Par. [7])
Regarding Claim 8, Wu does not explicitly disclose wherein: the aggregate of the contrastive loss function and the reconstruction loss function comprises a weighted combination of the contrastive loss function and the reconstruction loss function; and weights used in the weighted combination are included in the model parameters of the foundation model and are adjusted during the training to minimize the total loss function through an iterative back propagation process.
However, Truong discloses wherein: the aggregate of the contrastive loss function and the reconstruction loss function comprises a weighted combination of the contrastive loss function and the reconstruction loss function (See Par. [101], [107], [115], [126] of Truong for a reference to determining a combination weight of an updated loss function based on an updated similarity metric value, which is generated by applying the training mask on the original similarity metric); and weights used in the weighted combination are included in the model parameters of the foundation model and are adjusted during the training to minimize the total loss function through an iterative back propagation process (See Par. [101], [115], [124]-[126] and Fig. 7 of Truong for a reference to calculating the total loss function based on the comparison between the original [contrastive] loss function and the updated [reconstruction] loss function),
Thus, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of Truong to Wu. The motivation for combination would be to improve network’s performance, by improving the generation of machine learning models, through improving the security of sensitive data. (Truong; Par. [7])
Regarding Claim 9, the combination of Wu and Truong, specifically Wu discloses wherein training the plurality of task-specific models comprises at least one of:
freezing all model parameters of the foundation model during the training of the plurality of task- specific models (See Par. [103], [267] of Wu for a reference to that all parameters are frozen [No change is applied to model parameters] during the training of task models); fine-tuning all model parameters of the foundation model based on at least one task-specific prediction loss during the training of the plurality of task-specific models (See Par. [114], [116], [141]-[142], [196] of Wu for a reference to generating a respective series of training probe signals based on the obtained CI or TSCI associated with multiple events);
or during the training of the plurality of task-specific models: freezing model parameters of an upstream layer of the foundation model, and fine-tuning model parameters of a downstream layer of the foundation model, wherein each of the plurality of task-specific models is a downstream model compared to the foundation model.
Regarding Claim 11, the combination of Wu and Truong, specifically Wu discloses wherein performing the plurality of wireless sensing tasks comprises:
collecting real-time CI data from multiple wireless links for at least one task of the plurality of wireless sensing tasks (See Par. [144], [270] of Wu for a reference to determining real-time CI data from at least on sensing task of a plurality of sensing tasks); and for each task of the at least one task:
generating, using the foundation model, a plurality of feature maps each based on real- time CI data collected from a corresponding one of the multiple wireless links (See Par. [110], [196], [209]-[210] of Wu for a reference to that based on the obtained CI data over time, a feature map (environmental model) is created),
generating a fused feature map at least in part by fusing the plurality of feature maps along a subcarrier dimension or according to an index of each of the multiple wireless links (See Par. [110], [196], [209]-[210], [249] of Wu for a reference to that based on the obtained CI data over time, a fused feature map (environmental model) is generated), and inputting the fused feature map into a task-specific model corresponding to the task to generate a decision result for the task (See Par. [110], [209]-[210], [213], [249] of Wu for a reference to the environmental model is inputted the task training models to perform the sensing tasks).
Regarding Claim 12, the combination of Wu and Truong, specifically Wu discloses wherein performing the plurality of wireless sensing tasks comprises:
collecting real-time CI data from multiple wireless links for at least one task of the plurality of wireless sensing tasks; and for each task of the at least one task (See Par. [144], [270] of Wu for a reference to determining real-time CI data from at least on sensing task of a plurality of sensing tasks):
generating, using the foundation model, a plurality of feature maps each based on real- time CI data collected from a corresponding one of the multiple wireless links (See Par. [110], [196], [209]-[210], [213] of Wu for a reference to the environmental model is inputted the task training models to perform the sensing tasks),
inputting each of the plurality of feature maps into a task-specific model corresponding to the task to generate a candidate decision result for the task (See Par. [110], [196], [209]-[210] of Wu for a reference to that based on the obtained CI data over time, a feature map (environmental model) is created), and
fusing all candidate decision results generated for the task based on a fusion model to generate a final decision result for the task (See Par. [110], [209]-[210], [213], [249] of Wu for a reference to the fused environmental model is inputted the task training models to perform the sensing tasks).
Regarding Claim 13, the combination of Wu and Truong, specifically Wu discloses wherein:
the foundation model is trained based on self-supervised machine learning without labelled data (See Par. [140] of Wu for a reference to a projection for each CI data may be trained using a dimension reduction method, which comprises self-supervised learning); and
each of the plurality of task-specific models is trained based on labelled data (See Par. [119], [140], [177] of Wu for a reference to task-specific models are trained based on tagged/labelled data [Labelling Learning]).
Regarding Claim 14, the combination of Wu and Truong, specifically Wu discloses wherein:
the training dataset is generated by a local device and transmitted from the local device to a cloud server (See Par. [69], [82], [161] of Wu for a reference to Type 1 device [Transmitter] is a local device, while Type 2 device [Receiver] is a cloud server);
the foundation model and the plurality of task-specific models are trained by the cloud server (See Par. [114], [116], [141]-[142], [196] of Wu for a reference to generating a respective series of training probe signals based on the obtained CI or TSCI associated with multiple events); and
performing the plurality of wireless sensing tasks comprises:
collecting and processing real-time CI data by at least one local device to generate processed real-time CI data (See Par. [144], [270] of Wu for a reference to determining real-time CI data from at least on sensing task of a plurality of sensing tasks),
determining, by the at least one local device, whether a triggering event happens based on the processed real-time CI data (See Par. [90]-[91], [145] of Wu for a reference to determining by type 1 device whether a triggering event (e.g. motion detection) happens based on processing the real-time CI data),
in accordance with a determination that the triggering event happens, transmitting an immediate past portion of the processed real-time CI data within an immediate past time period from the at least one local device to the cloud server (See Par. [101], [119], [126], [188] of Wu for a reference to that if the triggering event happens, past portion of real-time CI data, over the past time interval, is transmitted to the Type 2 device), and
performing, by the cloud server, a wireless sensing task corresponding to the triggering event based on the immediate past portion of the processed real-time CI data using the foundation model and a task-specific model corresponding to the wireless sensing task (See Par. [88], [145], [166], [295] of Wu for a reference to the TSCI may be extracted from a motion sensing signal. Type 2 device performs smart sensing tasks, which are based on the trained characteristics [parameters] based on the extracted CI).
Regarding Claim 15, Wu discloses a system for wireless sensing (See Abstract and Fig. 6), comprising: at least one local device (See Fig. 9; Device 900) configured to:
obtain channel information (CI) data generated based on at least one wireless channel (See Par. [87], [94], [116], [147] and Fig. 11 of Wu for a reference to type 2 device receives a probe signal in a channel and obtains, at least, one channel information (CI) of the selected channel),
generate a training dataset based on the CI data (See Par. [114], [116], [141]-[142], [196] of Wu for a reference to generating a respective series of training probe signals based on the obtained CI or TSCI associated with multiple events), wherein the training dataset comprises: a plurality of CI pairs, original CI data and a mask (See Par. [141]-[144], [179], [194]-[196] of Wu for a reference to an event may be trained based on the projection and the training TSCI associated with the event. The projection maybe retrained using: the training TSCI, at least one current TSCI before re-training the projection and additional training TSCI of the transformation [masked] of the original TSCI); and
a cloud server (See Fig. 10; Device 1000) configured to:
train a foundation model using the training dataset at least in part by (See Par. [114], [116], [141]-[142], [196] of Wu for a reference to generating a respective series of training probe signals based on the obtained CI or TSCI associated with multiple events),
train a plurality of task-specific models (See Par. [114], [116], [140]-[142] of Wu for a reference to training a classifier of multiple events [Tasks] in a venue based on training TSCI associated with the multiple events. A projection for each CI may be trained based on the training TSCI); and
performing a plurality of wireless sensing tasks based on the foundation model and the plurality of task-specific models (See Par. [88], [145], [166], [295] of Wu for a reference to the TSCI may be extracted from a motion sensing signal. Type 2 device performs smart sensing tasks, which are based on the trained characteristics [parameters] based on the extracted CI), ), the performing comprises:
generating a feature map using the foundation model based on CI data collected in real-time (See Par. [110], [196], [209]-[210] of Wu for a reference to that based on the obtained CI data over time, a feature map (environmental model) is created), and
inputting the feature map into each of the plurality of task-specific models to perform a corresponding one of the plurality of wireless sensing tasks, respectively (See Par. [110], [196], [209]-[210], [213] of Wu for a reference to the environmental model is inputted the task training models to perform the sensing tasks).
Wu does not explicitly disclose determining a contrastive loss function based on a first similarity metric between CI data of each CI pair of the plurality of CI pairs in the training dataset, determining a reconstruction loss function based on a second similarity metric between the original CI data in the training dataset and predicted CI data generated based on the mask in the training dataset, determining a total loss function based on an aggregate of the contrastive loss function and the reconstruction loss function, and determining model parameters of the foundation model to minimize the total loss function; wherein: each of the plurality of task-specific models is a machine learning model concatenated after the foundation model, the foundation model is a machine learning model having a larger scale than each of the plurality of task-specific models
However, Truong discloses determining a contrastive loss function based on a first similarity metric between CI data of each CI pair of the plurality of CI pairs in the training dataset (See Par. [20], [101], [115], [124] of Truong for a reference to determining the loss function based on the similarity metric value between synthetic dataset [CI data] and normalized reference dataset [CI pair of the training dataset]),
determining a reconstruction loss function based on a second similarity metric between the original CI data in the training dataset and predicted CI data generated based on the mask in the training dataset (See Par. [101], [107], [115], [126] of Truong for a reference to determining an updated loss function based on an updated similarity metric value, which is generated by applying the training mask on the original similarity metric),
determining a total loss function based on an aggregate of the contrastive loss function and the reconstruction loss function (See Par. [101], [115], [124]-[126] and Fig. 7 of Truong for a reference to calculating the total loss function based on the comparison between the original [contrastive] loss function and the updated [reconstruction] loss function), and
determining model parameters of the foundation model to minimize the total loss function (See Par. [61]-[63], [71]-[75] and Fig. 3 of Truong for a reference to determining the training model through model optimizer that optimizes the loss function to the lowest levels); wherein: each of the plurality of task-specific models is a machine learning model concatenated after the foundation model (See Par. [62], [64], [79], [81] and Fig. 1-4 of Truong for a reference to generating a data model for a machine learning application, using synthetic data that can be generated using a synthetic dataset model, which can in turn be generated using actual data. A data model can be trained using computing resources using data provided by dataset), the foundation model is a machine learning model having a larger scale than each of the plurality of task-specific models (See Par. [62], , [81], [112], [126] and Fig. 1-4 of Truong for a reference to the scaling factor of the trained data model *Foundation model) is greater than each data model scaling factor)
Thus, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of Truong to Wu. The motivation for combination would be to improve network’s performance, by improving the generation of machine learning models, through improving the security of sensitive data. (Truong; Par. [7])
Regarding claim 16, the claim is interpreted and rejected for the same reason as set forth in claim 3.
Regarding Claim 17, the combination of Wu and Truong, specifically Wu discloses wherein:
processing the CI data comprises:
selecting subcarriers for at least one CI in the CI data to generate a same number of subcarriers for all CI in the CI data according to the standardized format (See Par. [87]-[89] of Wu for a reference to selecting a channel [Subcarrier] from a set of channels. At least, on CI may be obtained by type 2 device from the probe signals transmitted in the selected channel. The same number of channels are selected for all CI data); and
resampling each CI in the CI data to a predetermined temporal rate according to the standardized format (See Par. [119], [122], [177], [200] of Wu for a reference to resampling each CI data to a predetermined spatial-temporal normalized standard format); and
performing the data augmentation comprises at least one of:
adding random noise to the preprocessed CI (See Par. [210], [212], [251] and Fig. 9 of Wu for a reference to applying a noise cancellation method by adding noise to the extracted CSI data);
randomizing the selected subcarriers within a block (See Par. [87]-[89] of Wu for a reference to selecting a channel [Subcarrier] from a set of channels. At least, on CI may be obtained by type 2 device from the probe signals transmitted in the selected channel. The same number of channels are selected for all CI data);
performing a time scaling or a time warping on the preprocessed CI (See Par. [181], [305]-[307] of Wu for a reference to applying time scaling of preprocessed CI parameters);
simulating at least one environmental parameter related to multi-path change or occlusion (See Par. [148] of Wu for a reference to simulating multipath channel characteristics [Parameters]); and
normalizing amplitudes of the preprocessed CI to mitigate power variation (See Par. [192], [203], [251] of Wu for a reference to regulating signal amplitude of the processed CI to reduce CIR variation).
Regarding claim 18, the claim is interpreted and rejected for the same reason as set forth in claim 6.
Regarding claim 19, the claim is interpreted and rejected for the same reason as set forth in claim 7.
Regarding claim 20, the claim is interpreted and rejected for the same reason as set forth in claim 1, including a device for wireless sensing (See Fig. 9; Device 900), comprising: at least one processor (See Fig. 9; Processor 902); and at least one memory storing instructions (See Fig. 9; Memory 904).
Conclusion
5. The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Taralova (US. Pub. No. 2021/0027111 A1) discloses a method for determine whether the simulated environment is causes similar responses by the neural network.
Nadamuni Raghavan et al. (US. Pub. No. 2020/0302339 A1) discloses techniques are disclosed for training machine learning systems.
Lin et al. (US. Pub. No. 2019/00354802 A1) discloses systems, methods, and non-transitory computer readable media for utilizing a deep neural network-based model to identify similar digital images for query digital images.
6. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any extension fee pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the date of this final action.
7. Any inquiry concerning this communication from the examiner should be directed to RASHA FAYED whose telephone number is (571) 270-3804. The examiner can normally be reached on M-F 8:00AM-4:30PM.
If attempts to reach the examiner by telephone are unsuccessful, the supervisory Examiner, Un Cho can be reached on (571)272-7919. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of an application may be obtained from the Patent Application Information Retrieval (PAIR) system. Status information for published applications may be obtained from either Private PAIR or Public PAIR. Status information for unpublished applications is available through Private PAIR only. For more information about the PAIR system, see http://pair-direct.uspto.gov. Should you have questions on access to the Private PAIR system, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative or access to the automated information system, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/R.K.F/Examiner, Art Unit 2413
/UN C CHO/Supervisory Patent Examiner, Art Unit 2413