Patent Examiner Guidance - Examples
This page is under construction
Summary:
| Claim # | Topic | Eligibility | Rationale |
|---|---|---|---|
| Example 47 - AI based anomaly detection. | |||
| Claim 1 | An ASIC implementing an artificial neural network having an array of neurons and a plurality of synaptic circuits. | Eligible | Because ASICs are physical circuits, it is within a statutory class, and there is no judicial exception set forth in the claim. There is no mathematical ideas, mental processes, methods of organizing human activity, etc. |
| Claim 2 | A method of using an artificial neural network including training it using training data, then detecting anomalies using the trained neural network, analyzing and outputting the anomaly data. | Ineligible | Under Step 2A, prong 1, it is directed to a judicial exception of a mental process, and under Step 2A, prong 2, the claim as a whole fails to integrate the claim is directed to the judicial exception and the claim as a whole fails to integrate the judicial exception into a practical application. For example, the trained neural network is used generally to apply the agstract idea without placing any limits on how the trained neural network functions. Step 2B fails to save the claim because the claim as a whole does not amount to significantly more than the exception as the key elements are recited at a high level of generality and relate to well-understood, routine, and conventional activity. |
| Claim 3 | Adds steps of identifying anomalies associated with malicious network packets and dropping further packets from the source addressed based on same. | Eligible | Under step 2A prong 2, the invention reflects an improvement in the field of network intrusion detection and improved network security, and therefore integrates the abstract idea into a practical application. |
| Example 48 - Using AI to separate speech from background noise | |||
| Claim 1 | A method of using deep neural networks (DNN) to determine embedding vectors V of a spectrogram X of a mixed speech signal comprising speech from different sources. | Ineligible | The claim is directed to a judicial exception of a mathematical calculation. Under Prong 2 of Step 2A, the additional step of receiving the mixed speech signal “amounts to mere data gathering” and does not impose any other meaningful limits on the claims. The additional step of using deep neural network to determine embedding vectors lacks deatils of a particular DNN or how the DNN operates, and therefore only recites the idea of determining embedding vectors using a DNN. While the specification identifies an improvement to seech separation technology, the claim only requires determining embedding vectors and therefore does not reflect the improvement discussed in the disclosure. |
| Claim 2 | Adds steps of partitioning vectors V into clusters, applying masks to the clusters, synthesizing speach waveforms from the masked clusters, and combining the speech waveforms to generate a mixed signal excluding those from a target source, and transmitting the mixed speech signal. | Eligible | Steps f and g are directed to creating a new speech signal that does not contain extraneous speech signals from unwanted sources. The claimed invention therefore reflects this technical improvement by including these features. Therefore, the claim integrates the abstract idea into a practical application. |
| Claim 3 | A machine readable medium with instructions that cause a processor to perform operations of using a DNN to convert a time-frequency representation of a mixed speech signal into embeddings in a feature space, clustering the embeddings using a clustering algorithm, applying binary masks to obtain masked clusters, converting the masked clusters to obtain N separated speechc signals, and generating a sequence of words from the spectral features to produce a transcript of hte speech signal. | Eligible | |
| Example 48 - Using AI in Fibrosis Treatment | |||
| Claim 1 | Ineligible | ||
| Claim 2 | Eligible | ||
Navigation