Example 47

Claims that cover an artificial neural network, including a hardware claim and two method claims for using the network to improve security

Claim 1 (eligible) of the example recited an application specific integrated circuit (ASIC) for an artificial neural network. The claimed hardware comprises neurons organized in an array and the manner in which they were connected. The USPTO deemed this was directed to patent-eligible subject matter as it claimed a machine and/or manufacture, and the claim did not recite a judicial exception (i.e., an abstract idea, law of nature, or product of nature).

Claim 2 (ineligible) recited a method of using an artificial neural network. Several of the claimed steps described concepts that could be performed in the human mind, such as detecting and analyzing anomalies in information. The remaining steps failed to integrate the judicial exceptions into a practical application because they merely involved a generic computer, high-level data-gathering and output steps, and a statement of a desired outcome rather than a means of achieving the outcome.

Claim 3 (eligible) was also a method claim for using an artificial neural network. Although claim 3 recited an abstract idea (the same mental processes and mathematical steps as claim 2), it included three steps that showed a specific improvement in network security: they described how to take proactive measures to remediate danger by detecting source addresses associated with bad actors.

Example 48

Claims that cover a method of separating speech sounds from other sounds

Claim 1 (ineligible) involved receiving a mixed speech signal, converting it into a spectrogram, and using a neural network to determine embedding vectors. The USPTO explained that this effectively covers a mathematical formula, and is therefore directed to an abstract idea. Although the claim aimed to solve a specific problem—separating speech from different sources without prior knowledge of the number of speakers—Claim 1’s steps did not actually include the improvement discussed in the disclosure (partitioning the embedding vectors and synthesizing a mixed speech signal based on those clusters).

Claim 2 (eligible) was also a method claim, and although it included some mental process steps and a mathematical limitation, it added steps for creating new speech signals that achieve the claimed improvement and integrated the abstract idea into a practical application.

Claim 3 (eligible) was another method claim with steps that recited an abstract idea. But the USPTO opined that the “ordered combination of” receiving a signal, processing it to produce masked clusters, converting it to reduce masked clusters, and extracting spectral features to produce a transcript meant that the claim was directed to an improvement in existing speech-to-text technology and integrated the abstract idea into a practical application.

Example 49

Methods for treating fibrosis

Claim 1 (ineligible) covered a method for post-surgical fibrosis treatment that involved collecting and genotyping a patient sample, identifying the patient as high risk for inflammation based on a consolidated risk score, and administering an “appropriate treatment” to the patient. This method qualified as an abstract idea because the claimed step of identifying the patient as having a high risk of inflammation was a mental process. This step also recited a law of nature because it describes a naturally occurring relationship between a patient’s genes and gene-based risk factors. The claim failed to integrate the recited exception into a practical application because it did not describe the “appropriate treatment” to administer to the patient, and failed to recite an appropriately detailed inventive concept.

Claim 2 (eligible) recited the same steps as claim 1, but noted that the “appropriate treatment” is “Compound X eye drops.” Because the Compound X eyedrops were described as a “new anti-fibrotic drug,” the claimed requirement of administering the Compound integrated the claim into a practical application.