MUSCLE FREQUENCY FATIGUE, AND ASSOCIATED ALGORITHMS, SYSTEMS AND METHODS CROSS-REFERENCE TO RELATED APPLICATION This application claims the benefit of U.S. Patent Application No.63/215,250 filed June 25, 2021, the disclosure of which is hereby incorporated by reference in its entirety. BACKGROUND Market research and subject matter experts are showing that fatigue can make an athlete more susceptible to injury, and may, in fact, be one of the leading causes of injuries. Thus, there is a need to detect the onset of fatigue while an athlete is actively training, conducting practice, or participating in a live game. When a trainer (e.g., an athletic trainer or coach) detects signs of fatigue, the trainer can intervene to reduce the likelihood of fatigue-related injury. For example, when a trainer detects fatigue, the trainer may instruct the athlete to slow down or focus on technique. In addition, or in the alternative, the trainer may pull the athlete from a game or a practice session for rest and recovery. One of the challenges with detecting fatigue is that traditional methods of monitoring athletic performance, such as real-time heart rate monitoring, do not in and of themselves necessarily indicate when an athlete is fatigued. For example, an athlete experiencing lack of sleep, may still exhibit a normal or expected heart rate during a practice session, but may experience earlier onset of fatigue due to lack of sleep. As a result, the athlete may have a heightened susceptibility to fatigue-induced injury that is unknown to the trainer because the athlete's heart rate appears to be normal. Accordingly, there remains a need for cost effective algorithms, systems and methods that can detect early onset of fatigue that may not be readily detectable by real-time monitoring alone. BRIEF DESCRIPTION OF THE DRAWINGS The foregoing aspects and many of the attendant advantages of this invention will become more readily appreciated with reference to the following detailed description, when taken in conjunction with the accompanying drawings, where: Figure 1 is a diagram illustrating an analytics system configured in accordance with an embodiment of the present technology. Figure 2 is a diagram illustrating components of a system in accordance with an embodiment of the presently disclosed technology. Figure 3 is block diagram illustrating components of a system in accordance with an embodiment of the presently disclosed technology. Figures 4A and 4B are diagrams showing a measurement system in accordance with embodiments of the presently disclosed technology. Figure 5A is a graph of muscle amplitude versus time in accordance with an embodiment of the presently disclosed technology. Figure 5B is a graph of muscle amplitude versus frequency in accordance with an embodiment of the presently disclosed technology. Figure 6 is a graph of safe zones for exercise in accordance with another embodiment of the presently disclosed technology. Figure 7 is a graph illustrating measurement results in accordance with an embodiment of the presently disclosed technology. Figure 8 is a block flow diagram of an example process for determining and/or predicting fatigue in accordance with an embodiment of the presently disclosed technology. DETAILED DESCRIPTION A. System Overview Inventive technology is directed to a fatigue detection system that enables coaches, athletic trainers and/or the athletes themselves to determine when an athlete is experiencing fatigue, such as during a practice session or a live game. In general, an increase in muscle fatigue is characterized by increased amplitude and decreased frequency of neuromuscular activity of muscle fibers, for example, where both trends are present simultaneously. In some embodiments, the decreased frequency of muscle activity (e.g., the decreased frequency of muscle twitching) indicates fatigue. For example, when a user who normally starts her daily exercise at a frequency of 250 Hz for a particular muscle group, starts her next daily exercise at a frequency of, for instance, 180 Hz for the same muscle group, such lowered frequency may indicate fatigue before the exercise has started. Such fatigue may be caused by lack of sleep, stress, jetlag, etc. (collectively, environmental fatigue), and may be unaccompanied by an increase in the heart rate. In some embodiments, the athlete's uniform or other exercise clothing may be equipped with suitable sensors and/or data acquisition controllers that collect and interpret muscle activity data (e.g., muscle amplitude and frequency). Such sensors may measure electrical impulses of the muscles. Collected data may be processed to indicate starting points and trends that characterize muscle amplitude and/or frequency for one or more muscle groups of the user. When properly analyzed, the trends of the muscle amplitude and/or frequency can indicate whether the user is fatigued and to what extent. Analytics systems configured in accordance with various embodiments of the present technology, however, provide analytics that detect the early onset of fatigue that may not be readily detectable, for example, by heart rate monitoring. This, in turn, provides the opportunity for earlier intervention and corrective action to reduce the risk of fatigue- related injuries. Figure 1 is a diagram illustrating an analytics system 100 ("system 100") configured in accordance with an embodiment of the present technology. The system 100 includes a muscle activity tracker sub-system 102 ("muscle activity tracker 102") and a muscle monitoring sub-system 105 ("muscle monitor 105") that is worn by a user, such as an athlete or a user 111. The muscle monitor 105 may include an on-board controller 125 ("controller 125") and sensors 123 that can be integrated into the athlete's clothing (described in reference to FIGS.4A-4B), such as the athlete's shirt, pants, etc. The athlete's clothing and the integrated controller 125 and sensors 123 can be collectively referred to as "smart compression clothing." In operation, the controller 125 is configured to produce real-time or near real-time performance data ("real-time data") 107 during an exercise, live game, practice session, or conditioning. Analytics 110 includes muscle response (MR) data, like frequency and amplitude activity for different groups of muscles. In different embodiments, analytics 110 may include data related to orientation state (OS) of the user, acceleration of the user, activity state (AS) of the user, etc. The analytics 110 may be produced over an evaluation period of a certain duration (e.g., 1 hour, 30 minutes, 15 minutes, 5 minutes, etc.). As described below, the system 100 can use the analytics 110 to produce indications, warnings, and alarms that alert the user or the trainer when an athlete is fatigued or may soon become fatigued. Figure 2 is a diagram illustrating components of the system 100 in further detail. The system 100 illustrates interactions with multiple athletes, however, in other embodiments, the system may be focused on a single athlete. Furthermore, in different embodiments, the system 100 may include a subset of the illustrated components or additional components to those that are illustrated. The muscle monitor 105 shown in Figure 1 may be configured to communicate with one or more computing devices 206a-206f (collectively, computing devices 206) via a plurality of gateway devices 204 (e.g., gateway devices 204a-204d) positioned along monitoring region 227, such as a soccer-field, an athletic arena, gym, etc. The computing devices 206 are connected to one another via a network 208. The computing devices 206 are configured to receive, view, evaluate, store, and/or otherwise interact with data associated with the analytics 110 (Figure 1B). For example, intermediary or back-end server devices 206a and 206b can exchange and process communications over the network 208, store a central copy of data, globally update content, etc. Examples of well-known computing devices, systems, environments, and/or configurations 206a-206f that may be suitable for use with the technology include, but are not limited to, personal computers, server computers, handheld or laptop devices, cellular telephones, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, databases, distributed computing environments that include any of the above systems or devices, or the like. One or more computing devices 206a-206f can be configured to individually or collectively carry out the functions of the performance tracker 102 (Figure 1) for producing the analytics 110. In various embodiments, the computing devices 206 can process real-time data produced by one or more athletes 212-215 that are monitored in the monitoring region 227 of the gateways 204. As described below, the gateways 204 are configured to forward the data 107 (Figure 1) to the upstream computing devices 206 for processing (e.g., in real-time as a buffered stream and/or as intermitted transfers). B. Computing Devices Figure 3 is a block diagram illustrating components that can be incorporated into a computing device 301, such as one of the computing devices 206a-206f (Figure 2), the gateways 204 (Figure 2), and the muscle monitor 105 (Figure 1A). The computing device 301 includes input and output components 330. Input components can be configured to provide input to a processor such as CPU 331, notifying it of actions. The actions are typically mediated by a hardware controller that communicates according to one or more communication protocols. The input components 330 can include, for example, a mouse, a keyboard, a touchscreen, an infrared sensor, a touchpad, a pointer device, a camera- or image-based input device, a pointer, and/or a microphone. The CPU 331 can be a single processing unit or multiple processing units in a device or distributed across multiple devices. The CPU 331 can be coupled to other hardware components via, e.g., a bus, such as a PCI bus or SCSI bus. Other hardware components can include communication components 333, such as a wireless transceiver (e.g., a WiFi or Bluetooth transceiver) and/or a network card. Such communication components 332 can enable communication over wired or wireless (e.g., point-to point) connections with other devices. A network card can enable the computing device 301 to communicate over the network 208 (Figure 2) using, e.g., TCP/IP protocols. Additional hardware components may include other input/output components, including a display, a video card, audio card, USB, firewire, or other external components or devices, such as a camera, printer, CD-ROM drive, DVD drive, disk drive, Blu-Ray device, and/or speakers. The CPU 331 can have access to a memory 333. The memory 333 includes volatile and non-volatile components which may be writable or read-only. For example, the memory can comprise CPU registers, random access memory (RAM), read-only memory (ROM), and writable non-volatile memory, such as flash memory, hard drives, floppy disks, CDs, DVDs, magnetic storage devices, tape drives, device buffers, and so forth. The memory 333 stores programs and software in programming memory 334 and associated data (e.g., configuration data, settings, user options or preferences, etc.) in data memory 335. The programming memory 334 contains an operating system 336, local programs 337, and a basic input output system (BIOS) 338, all of which can be referred to collectively as general software. The operating system can include, for example, Microsoft Windows™, Apple iOS, Apple OS X, Linux, Android, and the like. The programming memory 334 also contains other programs and software 340 configured to perform various operations. The various programs and software can be configured to process the real-time data 107 of the athlete 111 (Figure 1A) and produce corresponding analytics, such as during the live session S1 (Figure 1B), as described in greater detail below. Those skilled in the art will appreciate that the components illustrated in the diagrams described above, and in each of the diagrams discussed below, may be altered in a variety of ways. C. Clothing and Sensors Figures 4A and 4B are diagrams showing a measurement system in accordance with embodiments of the presently disclosed technology. Referring to Figure 4A, the controller 125 can be embedded within the athlete's clothing, such as a shirt 445a and pants 445b (collectively "clothing 445"). In other embodiments, the controller 125 can be inserted into a pocket 443 in the user's clothing and/or attached using Velcro, snap, snap-fit buttons, zippers, etc. In some embodiments, the controller 125 can be removable from the clothing 445, such as for charging the controller 125. In other embodiments, the controller 125 can be permanently installed in the athlete's clothing 445. Referring to Figures 4A and 4B together, the controller 125 is operably coupled to muscle response sensors 423b that may be distributed over different muscle groups (e.g., pectoralis major, rectus abdominis, quadriceps femoris, biceps, triceps, deltoids, gastrocnemius, hamstring, and latissimus dorsi). The muscle response sensors 423b provide a measurement of the muscle activity during exercise. Amplitude and frequency of user's muscle response may be forwarded to the controller 125, and further to the computing devices 206 for data processing and display. A non-limiting example of the muscle response sensors 423b is an electromyography (EMG) sensor. The EMG sensors 423b can also be coupled to floating ground near the athlete's waist or hip. In some embodiments, the clothing 445 may also be equipped with electrocardiogram (ECG) sensors 423a, orientation sensors 423c (e.g., a gyroscope), and acceleration sensors 423d (e.g., an accelerometer). The sensors 423 can be connected to the controller 125 using thin, resilient flexible wires (not shown) and/or conductive thread (not shown) woven into the clothing 445. The gauge of the wire or thread can be selected to optimize signal integrity and/or reduce electrical impedance. The sensors 423a and 423b can include dry-surface electrodes distributed throughout the athlete's clothing 445 and positioned to make necessary skin contact beneath the clothing along predetermined locations of the body. The fit of the clothing can be selected to be sufficiently tight to provide continuous skin contact with the individual sensors, allowing for accurate readings, while still maintaining a high-level of comfort, comparable to that of traditional compression fit shirts, pants, and similar clothing. In various embodiments, the clothing 445 can be made from compressive fit materials, such as polyester and other materials (ex. Elastaine) for increased comfort and functionality. In some embodiments, the controller 125 and the sensors 423 can have sufficient durability and water-resistance so that they can be washed with the clothing 445 in a washing machine without causing damage. In these and other embodiments, the presence of the controller 125 and/or the sensors 423 within the clothing 445 may be virtually unnoticeable to the athlete. In one aspect of the technology, the sensors 423 can be positioned on the athlete's body without the use of tight and awkward fitting sensor bands. In the context of this application, the sensors 423 and the controller 125 are referred to as "wearable" components. In general, traditional sensor bands are typically uncomfortable for an athlete, and athletes can be reluctant to wear them. In additional or alternate embodiments, the muscle monitor 105 (Figure 1) can include a separate controller 446 worn on the athlete's pants 445b. The separate controller 446 can be similar to the controller 125 worn on the athlete's shirt 445a, and is connected to the individual sensors 423 located on the pants 445b. The separate controller 446 can be configured to communicate with the controller 125 and/or with the gateways 204 (Figure 2). D. Controller Communication In operation, the controller 125 of the muscle monitor 105 (Figure 1) is configured to process and packetize the data it receives from the sensors 423 (e.g., the muscle response sensors 423b). The controller 125 may broadcast the packetized data for detection by the gateway devices 204, which, in turn, forward the data to the performance tracker 105 (Figure 1A) to produce required analytics (e.g., frequency and amplitude of muscle activity). E. Fatigue Indication Figure 5A is a graph of muscle amplitude versus time in accordance with an embodiment of the presently disclosed technology. The horizontal axis indicates time in seconds, and the vertical axis indicates muscle amplitude in units of displacement (e.g., mm). The illustrated graph represents time series of the activity for a particular muscle group that was measured continuously by, for example, a muscle response sensor 423b. In this embodiment, time is divided into 3 periods: A, B and C. Generally, it may be difficult even for a trained observer to derive definitive conclusions about the trends in muscle amplitude and time from the illustrated time series of the muscle activity. A trained observer may conclude that amplitudes of muscle displacement are somewhat high during the period B, but even for the trained observer it may be hard to arrive to definitive conclusion regarding the muscle fatigue. The difficulties arise, at least in part, due to the multivariate nature of amplitude signals, being dependent on volitional and/or involuntary factors. For example, period B can correspond to a period of increased voluntary exertion (e.g., as part of a high intensity interval training program) or to a period of muscle fatigue with little to no indication in amplitude data. Without metadata labeling the amplitude data with additional information, the nature of period B can remain unclear. Advantageously, neurophysiological action of muscles and muscle groups can exhibit characteristic signs in frequency space that can inform fatigue analysis. For example, nerve impulse frequency (e.g., twitch rate) can vary with fatigue in an involuntary manner, making it possible to detect early fatigue indication from neuromuscular activity data. Therefore, in some embodiments, the muscle activity may be represented as a frequency vs. amplitude graph, as described below with reference to Figure 5B. Figure 5B is a graph of muscle amplitude versus frequency in accordance with an embodiment of the presently disclosed technology. The horizontal axis (ordinate) indicates frequency in Hz, and the vertical axis (abscissa) indicates muscle amplitude for a given muscle group in units of displacement. The two curves represent two periods of time during user's exercise or other activity. The illustrated curves are sometimes referred to as spectral graphs. Such spectral graphs may be obtained by transform operations including, but not limited to, Fourier transforms (e.g., a fast Fourier transform, short time Fourier transform, etc.). During the first period of time Δt ₁ , the muscle activity of the user (e.g., an athlete) occurs at a relatively high frequency f ₁ and at a relatively low amplitude A ₁ . Such a scenario indicates that the user is not fatigued yet, because a muscle group that is not fatigued generally exhibits higher impulse frequency and lower displacement amplitude. Therefore, an algorithm that compares frequencies and amplitudes of muscle activity may determine that a user is not fatigued at this first period of time Δt ₁ . Data related to the second period of time Δt ₂ are acquired after the user exercised for a duration of time or during a training or activity session during which the user is fatigued (e.g., after travelling, during a period of stress or lack of sleep, etc.). Here, the frequency of muscle activity decreases to f ₂ , while the amplitude of muscle activity increases to A ₂ . Such a scenario indicates that the user is getting fatigued or may already be fatigued, as indicated by a relatively low frequency and high amplitude of the observed muscle group. In some embodiments, an algorithm that compares frequencies and amplitudes of muscle activity may determine that the user is fatigued during the second period of time Δt ₂ . In other embodiments, based on the observed values of frequency and/or amplitude of the muscle group activity, a trainer may conclude whether an athlete is fatigued or not. In different embodiments, different values of amplitude/frequency represent determination thresholds that apply to different muscle groups. An example of the amplitude/frequency threshold determination is discussed with respect to Figure 6 below. In this way, fatigue can be quantified by controlling muscle activity data with respect to a control parameter, such as metadata or external load measurements generated using motion data, as described in more detail in reference to FIG.6. Fatigue quantification refers to techniques for generating a numerical value based at least in part on baseline corrected amplitude and/or frequency measurements generated during a given activity or task by the wearer 111. In an illustrative example, fatigue quantification can include determining a non-dimensional fatigue parameter (e.g., a scalar value between zero and one) based on a comparison of the amplitude of muscle activity to an internal muscle activity limit as a baseline, where the limit is determined from muscle activity data for the wearer that has been generated during an activity with a substantially equal external load value. The term “substantial” refers to a value that is not exactly equal to the stated value but for which deviation from the stated value results in little to no impact on the determination of the internal muscle activity limit, for example, as determined by analysis of models configured to determine the internal muscle activity limit for sensitivity to uncertainty in external load values. Figure 6 is a graph of safe zones for exercise in accordance with another embodiment of the presently disclosed technology. In some embodiments, an estimate of user's fatigue level can be made based on limited amount of data. For example, a relatively high amplitude of muscle activity in the very beginning of exercise may indicate that the user is already fatigued, because of, for example, lack of sleep or illness. Taking, for example, amplitude A ₁₀ as an acceptable threshold amplitude of the muscle activity at the beginning of the exercise, the user may be determined as being fatigued at the very onset of the activity if the measured muscle amplitude exceeds the amplitude A ₁₀ . Conversely, the starting muscle amplitude that is below the amplitude A ₁₀ may indicate an athlete or user that is not fatigued. Such amplitude thresholds may be different for different muscle groups, age of user, etc. Analogously, an acceptable starting frequency (e.g., f ₂₀ ) for the muscle group may also be determined for different muscle groups to establish a threshold frequency that determines whether the user is fatigued at the beginning of the exercise. Therefore, in at least some embodiments, even a single exercise (a pushup, a squat, etc.) may determine whether the user is fatigued. Each muscle and/or muscle group can be characterized by a different set of values, based at least in part on the tendency for different muscles to exhibit different neuro- physiological characteristics in response to different levels of exertion. Further, numerical values for amplitude and frequency can vary between different users based at least in part on general physical condition, genetic variation between individuals, as well as other factors that make a nominal threshold value for amplitude or frequency ineffective at quantifying and/or detecting fatigue. Instead, threshold values can be dynamic, defined based at least in part on individualized historical data, calibration exercises (e.g., employing timing analysis and other signal analysis techniques to identify muscle on-off sequences), and/or by controlling for external load measurements using motion data. In this context, external load refers to a measure derived from motion data that describes work by the user 111 on the environment, such as by moving the user 111 or moving an object. For example, external load can be based at least in part on the mass of the user 111, a total acceleration of the user 111 over a period of time, and a given distance traveled. In an example of a running activity, external load can be defined as the mathematical product of the mass of the user 111, the sum of acceleration sampled over the duration of the running activity, and the distance traveled by the user 111. It is understood that some activities can implicate a different derivation of external load, for example, where an activity includes jumping (e.g., basketball), relative motion (e.g., equestrian events), or gravity-aided motion (e.g., downhill skiing). To that end, algorithms of the present disclosure, as described in more detail in reference to FIG.8, can employ dynamic threshold values, such as A ₁₀ , f ₁₀ , f ₂₀ , or the like, which are defined for a given user and for a given activity. With dynamic thresholds, systems of the present disclosure can be configured to detect when user 111 is manifesting fatigue or the onset of fatigue on an individualized basis and can quantify the fatigue extent, rather than estimating fatigue based on standardized heuristics. Correlating muscle activity data with motion data and/or activity metadata permits time-series muscle data to be prepared for amplitude and frequency-domain analysis as part of fatigue determination/quantification. Nominal values for frequency measured from EMG signals can range from about 0 Hz to about 350 Hz, with average (e.g., mean or median) frequencies during periods of exertion of about 0 Hz to about 30 Hz, about 5 Hz to about 25 Hz, about 5 Hz to about 20 Hz, about 5 Hz to about 15 Hz, about 5 Hz to about 10 Hz. It is understood, however, that frequency measurements vary based at least in part on the muscle being measured, on the physical condition of the muscle being measured, as well as on fatigue state. In an illustrative example, a median frequency of an EMG signal measured at a right biceps muscle of user 111 can be about 86 Hz at rest, while a median frequency for the corresponding signal measured at a left biceps muscle of user 111 can be the same or different (e.g., about 78-80 Hz at rest). To that end, relative frequency measurements represent improved metrics for determining and quantifying fatigue state. For example, by differentiating frequency from a non-fatigued baseline (e.g., using calibration data and/or data from prior performances of the same activity), a frequency measurement can be controlled for variability introduced by factors other than fatigue state, such as activity type, general muscle condition, genetic and/or physiological factors, or the like. Nominal values for amplitude measured from EMG signals also vary widely, between about 30 μV and 30 mV, based at least in part on the muscle being measured, on the size of the muscle being measured, on the physical condition of the muscle being measured, as well as on fatigue state. For example, amplitude measurements can range from about 50 μV to about 350 μV for safe exertion. Above 350 μV, for example between about 400 μV and 800 μV, muscle motor unit damage can result. As with frequency measurements, however, variation in individual muscles and muscle groups, as well as the genetics, anatomy, physiology, or condition of the user 111 can vary nominal amplitude measurements between individuals and between activities. In an illustrative example, a better conditioned muscle can exhibit lower average amplitude measurements when controlling for the external load. For that reason, threshold values based on nominal amplitude (e.g., in terms of voltage measured) are less effective than measurements of amplitude based on a differentiation from a baseline. In some embodiments, amplitude measurements can be expressed as a difference, fraction, percentage, or ratio of an internal amplitude limit, termed a “maximum contraction.” Maximum contraction, in this context, refers to a peak average amplitude measurement (e.g., in microvolts) for a given activity that is a function of the condition of the muscle being measured. In this way, the term “maximum” does not refer to a global or perpetual maximum, but rather to an upper limit of the muscle being measured at the time and for the activity during which muscle activity is being monitored. In an illustrative example, data can be collected for the user 111 during a given exercise (e.g., an aerobic exercise such as a long-distance run) for which the user 111 maintains a given external load until experiencing severe fatigue. In another example, maximum contraction can be based on a fixed position exercise such as weightlifting, for which metadata is used to control for variability in muscle activity data (e.g., metadata labels correlated with EMG data timestamps). Maximum contraction does not refer to an absolute limit of the given muscle, but rather to a limit at the current condition of the given muscle that can increase or decrease based at least in part on the health of the user 111, the physical condition of the user 111, the intensity of the activity, or other factors. In some embodiments, threshold values, such as A ₁₀ , are defined in relation to the maximum contraction. As part of a training program, the ratio of the maximum contraction can be determined based at least in part on a goal of improving the condition of the muscle, which corresponds to a reduction in the maximum contraction over time. In this way, threshold values can be defined as about 10%, about 20%, about 30%, about 40%, about 50%, about 60%, about 70%, about 80%, about 90% of maximum contraction, including fractions and interpolations thereof. For a lower threshold value, muscle fatigue is reduced by indicating fatigue sooner during a series of repetitions at a cost of more incremental improvement in muscle condition. For a higher threshold value, fatigue indication is delayed, coincident with an increased risk of muscle damage. For that reason, fixed threshold values represent a balance of competing objectives, namely training efficacy and risk avoidance. In some embodiments, threshold values, such as A ₁₀ , f ₁₀ , f ₂₀ , are defined by monitoring a rate of change in amplitude and/or frequency (e.g., as a percentage of maximum contraction). In an illustrative example, mean or median amplitude can increase over the course of an activity as the user 111 begins to manifest fatigue up to a point where severe fatigue onset occurs. In some cases, the onset of severe fatigue is observable by a significant increase in amplitude measurement (e.g., detectable as a step change in percent of mass contraction data). As illustrated in FIG. 7, the mean amplitude in sprint number 11 is significantly higher than in sprint number 10. Within a single activity, such as a long-distance run, a step-change can be observable as well, measured as a percentage of max contraction or as baseline-corrected nominal amplitude. While such changes are also detectable in frequency data, the magnitude of the change can be less significant. A zone of the graph where both the frequency and amplitude of a given muscle group are within the acceptable starting values is marked as "safe zone 1." In this zone, the starting frequency of the muscle group is higher than f20, while the starting amplitude is below A ₁₀ . In general, starting frequency of the muscle activity tends to be more determinative about user's fatigue than the starting amplitude. Therefore, in the illustrated graph, the zone that is to the right of the frequency f ₂₀ , but above the amplitude A ₁₀ (safe zone 2) may still indicate an acceptable level of fatigue for the user to proceed. In different embodiments, the changes in amplitude and/or frequency of the muscle group may be observed over a single practice or over a course of several days or many days. A sample observation in the muscle group amplitude/frequency is explained with reference to Figure 7 below. Figure 7 is a graph illustrating measurement results in accordance with an embodiment of the presently disclosed technology. The horizontal axis shows time in seconds. The vertical axis shows muscle frequency (lower section of the graph) and muscle amplitude (upper section of the graph) as measured by sample sensors 423b attached to subject's quad muscle group. The measurement results are divided into 12 trial sprints, each lasting about 90 seconds. A short rest period comparable to a duration of the sprint was administered after each sprint. The test subject was an experienced sprinter. The test results may be roughly divided into two groups. The first group of about 8 sprints are characterized by a steady increase of the muscle amplitude and a steady decrease in the muscle frequency. Such trends of muscle performance are indicative of fatigue increase. As the test subject becomes even more fatigued, a severe fatigue zone is reached starting at about sprint number 9. Within the severe fatigue zone, the muscle activity is characterized by severe increases in the amplitude (trendline 510). Additionally, the frequency of the muscle activity becomes less coherent (spread over a larger range of frequencies) as different groups of muscle fibers start operating at different frequencies (trendline 520). When the sprinter becomes fatigued, the frequency of the muscle activity after the sprint does not necessarily recover to the level that characterized the first several sprints, as can be observed by comparing the frequencies during the rest time after the initial sprints vs. after the later sprints. Based on the above measurement results, a computer algorithm or a person of ordinary skill, for example a trainer, would during the training recognize that any sprints beyond about 7-8 initial ones are counterproductive for the subject. FIG. 8 is a block flow diagram of an example process 800 for determining and/or predicting fatigue, in accordance with embodiments of the disclosure. As with FIG. 2, operations of example process 200 can be performed by one or more components of example system 100, and can be parallelized, distributed, repeated, reordered, or omitted. In some embodiments, the operations of example process 800 are performed by a single component of example system 100, such as computing device(s) 206 (e.g., server(s)). As such, example process 800 represents a set of operations performed by a computer in accordance with instructions encoded in software. At operation 805, example process 800 includes one or more initialization operations executed by constituent components of example system 100. In some embodiments, operation 805 can also include detecting sensors 423 included as part of a wearable sensor platform, for example, based at least in part on a number and/or configuration of sensors 423, metadata identifying sensors 423, or the like. In some embodiments, initialization can include internal calibration operations for sensors 423, such as initializing inertial measurement units or background/baseline definition for EMG sensors. In an illustrative example, operation 805 can include one or more prescribed motions by wearer to provide a standard initiation sequence of movements, such that the wearable sensor platform can be calibrated in reference to calibration data (e.g., stored in computing device(s) 206) and/or sensors 423 can be identified from preliminary sensor data (e.g., muscle groups can be detected from initial neuromuscular data). Advantageously, such initialization operations using sensors 423 can facilitate subsequent operations of example process 800, such as generating muscle activity data and/or monitoring fatigue based at least in part on an identification of the monitored muscle group. At operation 810, example process 800 includes generating muscle activity data. Generating muscle activity data can include constituent operations for measuring and storing EMG signal data by computing device(s) 206, receiving data communicated from another electronic device (e.g., sensor(s) 423), or the like. As such, generating muscle activity data by computing device(s) 206 can include receiving neuromuscular activity data and/or motion data from one or more sensors 423, describing the amplitude of neuromuscular signals as a function of time, and/or translation, rotation, and orientation of sensor(s) 423 as a function of time, recorded during performance of a task by the user 111 (e.g., an exercise, an athletic activity, a physical performance, etc.). As described in reference to FIGS.5A-7, muscle activity data can be generated by sensors 423 in electrical contact with or more muscles or muscle groups of user 111. Muscle activity data can be used to detect early indications of fatigue onset and to generate a fatigue prediction, based at least in part on time-series amplitude data and/or frequency-space data derived from EMG signals. To that end, at operations 815 and 820, example process 800 includes processing muscle activity data to generate frequency-domain data and time- domain amplitude data, respectively. As described in more detail in reference to FIGs.5A- 7, frequency-domain data can indicate whether user 111 is indicating fatigue by slowing down (e.g., decreasing frequency of repetitive motions) and/or manifesting fast-twitch muscle fatigue (decreasing frequency of nerve impulses). Similarly, amplitude data can reveal whether nerve impulses of a given muscle are increasing the average peak impulse amplitude for the same work, which can indicate fatigue. To that end, operations 815 and 820 represent data preparation operations undertaken using muscle activity data generated during motion by user 111 while wearing sensor(s) 423, to prepare data for analysis using one or more logical operations to predict or determine a fatigue state of user 111. At operations 825-840, example process 800 includes various decisions that can be taken based at least in part on muscle activity data, frequency data, and/or amplitude data. For example, at operation 825, frequency domain data can indicate whether the average impulse frequency satisfies a safe frequency, as described in more detail in reference to FIG. 6. In this way, where the average impulse frequency is below the threshold value, example process 800 can proceed to a generating a fatigue prediction at operation 845. Similarly, at operation 830, where the average peak amplitude exceeds a threshold value, example process 800 can proceed to operation 845. At operations 835 and 840, example process 800 includes examining rate of change parameters of frequency-domain and amplitude data. For example, operation 835 includes determining whether muscle activity data are exhibiting a decrease in average impulse frequency over time (e.g., by determining a derivative value for average impulse frequency). In such cases, user 111 can be manifesting fatigue detectable in frequency- domain data that is invisible in peak amplitude data. Similarly, operation 840 includes determining whether the average peak amplitude is increasing with time, indicative of muscle fatigue onset. Operations 825-840 represent a set of algorithmic decisions that represent a Boolean OR operation, corresponding to safe zone 1 in FIG.6. To that end, where average impulse frequency, average peak amplitude, rate of change of average impulse frequency, OR rate of change of average peak amplitude fail to satisfy a given threshold (e.g., negative slope, positive slope, f ₂₀ , A ₁₀ , etc.), example process 800 can include generating a fatigue prediction. To that end, at operation 845, example process 800 includes generating a fatigue prediction based at least in part on the processing of muscle activity data, frequency data, and time-domain dynamics data. In some embodiments, operations 825-840 represent a different Boolean operation, represented by the combined safety zones 1-3. For example, operation 845 can include logic such that fatigue onset is indicated where frequency is less than f20 and amplitude is greater than A10, or where frequency is less than f20 and amplitude is less than A10. In this way, logic encoded in software can be configured to interpret frequency and amplitude data as part of detecting fatigue. Where user 111 does not indicate or manifest fatigue onset, example process 800 can include monitoring muscle activity by repeating muscle activity data generation and processing operations. In this way, user 111 can be monitored on an ongoing basis during an exercise, task, performance, or other activity. In some embodiments, fatigue prediction is based at least in part on one or more outcomes of a rules-based model to which the muscle activity data are inputs. For example, a model can include functions that input muscle activity data and generate numerical inputs to a fatigue prediction model, such that the value returned by the fatigue prediction model can indicate a likelihood that user 111 is already fatigued or will become fatigued. In some embodiments, generating a fatigue prediction can include Boolean logic, whereby a fatigue prediction model inputs derived values from motion, frequency, and derivative data, and outputs a true or false value as a function of time describing whether user 111 is fatigued or whether future fatigue is indicated. At operation 850, example process 800 includes termination operations for monitoring and predicting fatigue onset. In some embodiments, operation 850 can includes scheduling a session for user 111 with a massage therapist, a physical therapist, or other service provider to assess and correct fatigue. In some embodiments, operation 850 includes generating fatigue notifications to be pushed to computing device(s) 206 of user 111, such as a smart phone or other electronic device, for example, as part of an interactive user environment through which user 111 can access and/or review muscle activity data and fatigue notifications as part of a training system. In this way, operation 850 can include adding or removing exercises, activities, or athletic performances from user’s 111 training or competition schedule, but can also include adding consultations, appointments, reviews, or other sessions to a schedule of user 111, as part of a holistic safety protocol aimed at preventing and treating muscular stress and fatigue. While various advantages associated with some embodiments of the disclosure have been described above, in the claims, and the Appendix in the context of those embodiments, other embodiments may also exhibit such advantages, and not all embodiments need necessarily exhibit such advantages to fall within the scope of the invention. For example, while various embodiments are described in the context of an athlete (e.g., a professional or collegiate athlete), in some embodiments users of the system can include novice or intermediate users, such as users, trainers, and coaches associated with a high school sports team, an athletic center, a professional gym, physical therapist, etc. Accordingly, the disclosure can encompass other embodiments not expressly shown or described herein.