Beyond Moreer

WTM2101 chips have howling suppression and noisy cancellation functions, especially in noisy cancellation. The latency is less than 10ms，PESQ achieves 3.12 and STOI achieves 0.97. The noisy algorithm is significantly better than the traditional algorithm.

It's worth noting that obtaining an accurate and representative reference sample is crucial for effective noise cancellation. If the reference sample contains any desired audio signal or if it doesn't accurately represent the background noise, the cancellation process may not be as successful in reducing unwanted noise.

WTM2101 chip in computing in memory structure holds high AI computing power in 100X and low power consumption in 5uA-3mA. WTM2101 uses NN algorithm and traditional algorithm. Compared with a single traditional algorithm, the WTM2101 chip could accurately recognize different noises and input the transparent human voice.

Noise cancellation in handheld transceivers typically involves the use of digital signal processing algorithms. These algorithms analyze the received audio signal and identify the components that are likely to be noise. Then, they generate an anti-noise signal that is out of phase with the identified noise, effectively canceling it out when combined with the original signal.

- More easily achieve the customized requirements for ordering different speech languages and keywords. On the other hand, it could be more accurate to recognize the keywords in low latency. WTM2101 fulfilled the request in the applications, due to DNN-HCLG models and NN algorithms.

- Smart home owners can input order words to control the home devices. The behaviour for accuracy and speed is direct distinction for each speech recognition chip solution, WTM2101 can achieve microwatt power consumption and 50 Gops AI computing。

- With computing in memory structure, the memory wall has been broken, computing and memory exists in one unit, improving AI computing and lower power consumption. WTM2101 is the CIM chip which has been applied to intelligent voice and intelligent health. Speech recognition is the significant function in WTM2101 and has been widely used in smart wearable devices.

Computing in memory chips can significantly reduce power consumption, improve battery life, and are suitable for scenarios such as mobile devices that require long-term use. At the same time, it can also reduce power losses.

Computing in memory chips can be applied to the field of artificial intelligence (AI) to accelerate the training and inference of deep learning algorithms, thereby improving AI performance and efficiency. WTM2101 from Witmem can be applied to accelerate deep learning algorithms, improving the performance and efficiency of AI applications.

- Computing in memory chips can also be applied to data centers and cloud computing environments for storing and processing large-scale data, thereby improving data center efficiency and energy utilization.

The computing in memory architecture works very well in handling large-scale machine learning models, as these models typically require a large amount of storage and computing resources. However, for some smaller models, computing in memory architecture may not be the optimal choice, as it may increase system complexity and cost. When choosing computing in memory architecture, factors such as...

The computing in memory architecture can reduce energy consumption by using storage and computing resources more efficiently. Because the need for data transfer is reduced, computing in memory architecture can reduce memory and disk access during training and inference, thereby reducing the energy consumption of the system. This is an important advantage for organizations that care about energy...