SemusiSDK Accuracy - A Comparative Analysis

Semusi has been developing cutting-edge Machine Learning solutions with its SemusiSDK for Human Activity Recognition, Demographics recognition, Place labeling and Interest detection. As our systems have matured, we are now benchmarking our solutions against popular Android applications in the market.

This benchmark is against “Moves”, a popular pedometer applications that measures the steps taken by a user and the duration the user has been walking. We asked 6 people - 3 females and 3 male in their mid-20's, to walk 500 steps. The people were between 5 feet 3 inches to 6 feet in height and between 60-80kg in weight. We used 4 high- and low-end devices held in hand, which took 4.47 minutes of walking time. The GPS in all devices was turned off. The devices used were -

1. Samsung S4
2. Samsung S2
3. Moto G
4. Micromax A21

Summary:

Counting Steps -

As we can see from the graphs, SemusiSDK performs with 63.39% better accuracy on average in counting steps than Moves. SemusiSDK logged 323.5 steps on average, as against 182.58 by Moves.

SemusiSDK is 63.39% more accurate in the best case when using Micromax A21. SemusiSDK logged 285 steps on average, as against 104.33 by Moves.

SemusiSDK performs 59.80% better for females and 25.48% better for males than Moves for counting steps. SemusiSDK logged 340.83 steps on average for females and 306.17 for males, as against 137 and 228.17 respectively by Moves.

Counting Walking Duration -

As can be seen, SemusiSDK is 3.67x better on average for calculating minutes walked than Moves. SemusiSDK logged 3.21 minutes on average, as against 0.69 by Moves.

In the best case, SemusiSDK is 9.18x better when calculating minutes walked than Moves on Samsung Galaxy S4. SemusiSDK logged 3.39 minutes on average, as against 0.33 by Moves.

As in the plots, we can see that SemusiSDK 5.38x better for calculating minutes walked for females and 2.56x better for males than Moves. SemusiSDK logged 3.45 minutes for females and 2.97 minutes for males on average, as against 0.54 and 0.83 respectively by Moves.

 Figure 1: Average steps calculated by SemusiSDK and Moves, and the actual steps walked

 Figure 2: Average steps using Samsung Galaxy S4 over 3 Females, 3 Males

 Figure 3: Average steps using Samsung Galaxy S2 over 3 Females, 3 Males

 Figure 4: Average steps using Moto G over 3 Females, 3 Males

 Figure 5: Average steps using Micromax A21 over 3 Females, 3 Males

 Figure 6: Average steps calculated by SemusiSDK and Moves for 3 Females

 Figure 7: Average steps calculated by SemusiSDK and Moves for 3 Males

 Figure 8: Average minutes walked as calculated by SemusiSDK and Moves, and the actual steps walked

 Figure 9: Average minutes walked using Samsung Galaxy S4 over 3 Females, 3 Males

 Figure 10: Average minutes walked using Samsung Galaxy S2 over 3 Females, 3 Males

 Figure 11: Average minutes walked using Moto G over 3 Females, 3 Males

 Figure 12: Average minutes walked using Micromax A21 over 3 Females, 3 Males

 Figure 13: Average minutes walked by SemusiSDK and Moves for 3 Females

Figure 14: Average minutes walked by SemusiSDK and Moves for 3 Males

Table 1

The actual logged data is as below.

Actual Steps – 500

Actual Minutes – 4.47

User	Device	SemusiSDK Recorded Steps	SemusiSDK Recorded Minutes	Moves Recorded Steps	Moves Recorded Minutes
Female 1	Samsung Galaxy S4	507	3.37	383	0
	Samsung Galaxy S2	259	2.92	0	0
	Moto G	53	1.17	60	0
	Micromax A21	310	3.9	5	0
Female 2	Samsung Galaxy S4	794	4.9	463	0
	Samsung Galaxy S2	380	4.5	0	0
	Moto G	281	5.22	18	0
	Micromax A21	310	3.9	5	0
Female 3	Samsung Galaxy S4	434	2.78	440	0
	Samsung Galaxy S2	295	2.81	270	2
	Moto G	265	3.32	0	0
	Micromax A21	202	2.55	0	4.47
Male 1	Samsung Galaxy S4	335	3.05	319	0
	Samsung Galaxy S2	273	2.6	101	2
	Moto G	278	3.29	0	0
	Micromax A21	225	2.55	0	0
Male 2	Samsung Galaxy S4	554	3.38	469	2
	Samsung Galaxy S2	232	2.78	163	0
	Moto G	293	3.94	176	0
	Micromax A21	329	2.69	435	0
Male 3	Samsung Galaxy S4	274	2.88	390	0
	Samsung Galaxy S2	290	2.58	270	2
	Moto G	257	3.31	234	4
	Micromax A21	334	2.5995	181	0

#BeContextAware

Drumming the #BeContextAware beat !!
E27 is a leading startup ecosystem publisher in Asia.  Semusi made the short list to showcase our technology at Echelon 2014 in Singapore.  Had a great day - talking to many mobile developers who were intrigued and excited about the new ways to develop exciting engaging context aware apps.  

We demoed Semusi API's, samples apps, customer apps and our Semusi Smart Targeting beta.

 

Marketing in the Age of Context

"In marketing, content is king but context rules. Context rules because it makes the experience relevant”

Last Sept (2013) I attended the Mobile Marketing Forum event in India where I gave a give a talk on 'conquering mobile advertising holy grail: Context-Awareness matters'. The premise of my talk was that mobile marketing needed to become more engaging, relevant and smarter! In a world of 'one size fits all' marketing, I urged mobile marketers to start to think about realizing the full potential of mobile marketing and to deliver on the promise of engaging 'right message, right place, right time' sometimes referred to as 'marketing in the moment'. A prerequisite to deliver on this promise, was for marketers to start thinking about a user's context, succinctly referred to as 'context awareness'.

To illustrate the potential power of context in marketing I previewed Semusi work on activity context and demoed how mobile banner ads could change in real time based on the user activity - sitting, standing etc

In Nov (2013) Robert Scoble & Shel Israel‘s launched their high anticipated book 'Age of context'. The authors did a great job in making the case that the combination of 5 forces; Mobile devices, data, sensors, social and location will give rise to a new era where context will play a central role.

“As discrete entities, each force is already a part of life. Together they have created the conditions for an unstoppable perfect storm of epic proportion"

The great thing about a high profile book launch is that its gets everyone talking. Over the past few months I have enjoyed talking to many marketers about the relevance of context in marketing.  In these discussions, marketers often ask for comparisons with native advertising and web based contextual marketing where content is used to understand context. We believe content consumption for a user is at best one signal of user context. To truly understand user context requires deciphering multiple signals - content, location, sensors to name a few.

Semusi is 100% focused on fusing multiple signals to provide an easy to use context awareness platform for marketers and developers. In my interactions with marketers,  many of them remarked that whilst understanding the activity context of users in exciting, it would be great to understand the Demographic Context for users; Gender, Weight, Heights...using sensors.  Well the Semusi team got to work ....today  at the Mobile Marketing Forum in Singapore (May 2014) I will be demoing Demographic Context in action! I look forward to meeting marketers and brainstorming how marketing will change in the Age of Context.

The Places in our lives

This post is about how “Places” as a semantic notion is more powerful in some ways than “Location” as a geophysical notion. 

Take a look at how you interact in offline or physical world. Do you say “I am at A-41 Corenthum, Tower A….  (or equivalently I am at 28.6272, 77.3735)“ or do you say “I am at office”

The former is a way of expressing Location and the latter is a way of expressing a Place.

A lot of services over time have been built using LBS (or Location Based Services) but unsurprisingly none of them have been successful. Whilst people have given several reasons, one of the reasons could be that just because you are in a particular location doesn’t mean that you are interested in services for that location, even though the likelihood is there.  For e.g. just because you are in the vicinity of a Pizza Hut doesn’t mean you are interested in buying a Pizza. You could be passing that location regularly everyday on your commute.

So what really matters is the significant locations in your life, like Office, Home or Transportation (if you consider a car or a bus as a Place)

What about business establishments that want to target users who are in vicinity – the poster child of LBS? Well, they should look at other signals – for example is the user found in other “Places” like the current business? Is the user in the vicinity of a Place that serves food and its lunchtime?

I think that the discussion has to move beyond Location to Place, Time, Demographics, and other aspects of a User to enable these services to be meaningful and succeed.

Use Cases for Mobile Context Aware Systems

In the previous post I had discussed why are context aware systems important now (link)

Its natural to ask that given this importance, how can these systems be useful.

Well, before we answer that question we have to define what "context" is

We like to define context by breaking it down into 4 parts (Who, What, Where, When)

Who: This is your identity and relates to your gender, age, weight, height, background, profession etc
What: This looks at what your current activity, ambience or interests are.
Where: This is your location (not just a lat, long) but also what type of taxonomy can we attach to that location for e.g. home, office, commercial, residential, restaurant, etc and also add more attributes around that location.
When: This tells you what time of the day, week, month, year etc is relevant to the current context.

So, given this definition of context we can define that a context can be used (i.e. personalised) as follows

WHEN a WHO is doing WHAT at WHERE then do ACTION

The use cases can then be derived as following

Mobile Marketing - Context Awareness Matters

In this presentation at MMF 2013 Dilip gives a glimpse of what is possible if you are aware of the "activity" context of the user and why context matters in advertisement.

Conquering mobile advertising holy grail: Context-Awareness matters - Dilip Mistry SEMUSI from Semusi

Singapore Quantified Self

Watch Dilip Mistry enthrall the audience with his tricks at the Singapore QS meet :-)